Materials Chemistry A: Journal of

Journal of
Materials Chemistry A
View Article Online
Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.
REVIEW
Cite this: J. Mater. Chem. A, 2015, 3,

8943
View Journal | View Issue
Perovskite-based solar cells: impact of morphology

and device architecture on device performance
Teddy Salim,a Shuangyong Sun,a Yuichiro Abe,abc Anurag Krishna,abc
Andrew C. Grimsdalea and Yeng Ming Lam*a
Organicinorganic metal halide perovskites have recently shown great potential for application in solar cells
with excitingly high performances with an up-to-date NREL-certied record eciency of 20.1%. This family
of materials has demonstrated considerable prospects in achieving eciencies comparable to or even
better than those of thin lm solar cells. The remarkable performances thus far seem not to be limited to
any specic device architecture. Both mesoscopic and planar cells showed good device performance
and this eventually leads to the inevitable comparison between both architectures. Regardless of device
architecture, device performance is highly dependent on the lm morphology. The factors inuencing
Received 1st October 2014

Accepted 17th December 2014
the lm morphology such as the deposition method, material composition, additives and lm treatment
will be discussed extensively in this review. The key to obtaining good-quality lm morphology and
hence performance is to essentially lower the energy barrier for nucleation and to promote uniform
growth of the perovskite crystals. A comparison of the material selection for various layers as well as
DOI: 10.1039/c4ta05226a
their corresponding impact on the perovskite lm and device behavior in both device architectures will
www.rsc.org/MaterialsA
be presented.
1. Introduction
The use of organicinorganic metal halide perovskites in
optoelectronic devices, such as eld eect transistors and
optical devices, is not new.13 That being said, the viability of
this class of materials for solar-to-electrical power conversion
was only demonstrated in 2009 by Kojima et al.4 Ever since
then, there has been no looking back for these perovskite
based solar cells. The power conversion eciency (PCE) of
solar cells based on methylammonium lead halide (CH3NH3PbX3 where X can be I, Br or Cl) has increased dramatically from the initial 3.9% to more than 19%5 in a short span
of less than 5 years. In comparison, multicomponent chalcogenide thin-lm solar cells, such as CuInxGa1xSe2, took 35
years to achieve a similar improvement in eciency. More
interestingly, the good performance so far seems not to be
conned to a single type of ideal architecture. Both mesoscopic solar cells (similar to solid state dye-sensitized solar
cells) and planar heterojunction (PHJ) solar cells have shown
impressive eciencies. On top of this, there is a wide range of
techniques, ranging from solution to vacuum based, that can
be used for the fabrication of highly ecient devices with
a
School of Materials Science and Engineering, Nanyang Technological University, 50

Nanyang Avenue, 639798, Singapore. E-mail: ymlam@ntu.edu.sg
Energy Research Institute@NTU (ERI@N), Nanyang Technological University,

Research Techno Plaza, Singapore 637553 637141, Singapore
c
Energy Research Institute@NTU (ERI@N), Interdisciplinary Graduate School,
Nanyang Technological University, Singapore 637553, Singapore
This journal is The Royal Society of Chemistry 2015
either architecture. It is precisely the robustness of the

devices demonstrated so far that has convinced both scientic and industrial communities that the eciencies can be
further improved if there is a better understanding of how
optimal processing conditions are related to the device
architectures and how this in turn aects the device
performance.
Organicinorganic metal halide perovskites, specically
methylammonium lead iodide (CH3NH3PbI3), were initially
utilized as sensitizers for liquid-state dye-sensitized solar cells
in the form of nanoparticles decorating the surface of a mesoporous metal oxide lm.4,6 Though reasonable PCE was achieved when coupled with a iodide/triiodide liquid electrolyte,
the stability of the cells was poor due to the dissolution of the
perovskite layer. A breakthrough moment for these mesoscopic
solar cells came in 2012 when the perovskite was coupled with
organic small molecule 2,20 ,7,70 -tetrakis(N,N-di-p-methoxyphenyl-amine)-9,90 -spirobiuorene (spiro-OMeTAD) as the hole
transport material (HTM) analogous to a solid state dye sensitized cell (DSSC).7,8 Kim et al. showed CH3NH3PbI3 solar cells
with a titanium oxide (TiO2) scaold could result in a PCE of up
to 9.7%.8 Transient absorption studies of these cells revealed
that holes were injected from CH3NH3PbI3 into spiro-OMeTAD
following the electron transfer from the perovskite phase to the
mesoscopic TiO2 lm. Since the TiO2 scaold participates in the
photoconversion process, it can be considered an active
component in the perovskite solar cell; hence we designate TiO2
and metal oxides with similar functions as active scaolds.
J. Mater. Chem. A, 2015, 3, 89438969 | 8943
View Article Online
Journal of Materials Chemistry A
During the same period, Lee et al. demonstrated a variation

from these active scaold mesoscopic cells. A mixed halide
perovskite (CH3NH3PbI3xClx) was deposited on an insulating
aluminum oxide (Al2O3) scaold and this was coupled with
spiro-OMeTAD. With this conguration, an impressive PCE of
10.9% could be achieved. These types of solar cells are now
more commonly known as meso-superstructured solar cells
(MSSCs).7 In MSSCs, the methyl ammonium lead halide acted
as both sensitizer and electron transport material (ETM). The
metal oxides in MSSCs only function as a scaold and electrons
could not be injected into the insulating Al2O3 layer due to its
wide bandgap. Since these insulating mesoporous metal oxide
lms do not participate actively in the charge transport and only
act purely as a supporting scaold for perovskites, they are
considered as a passive scaold (or inactive scaold). In
general, regardless of whether the scaold is active or passive,
thicker and more uniform perovskite lms can be deposited on
these scaolds. This in turn translates to higher light absorption and improved electrical properties for these cells.
The other organicinorganic metal halide-based solar cell
device architecture that coexists symbiotically with the mesoscopic structure discussed above is the planar heterojunction
(PHJ) architecture. In this conguration, a semiconducting
absorber is typically sandwiched between two selective contacts
in the absence of mesoporous (or nanoporous) scaold. This
type of architecture possesses a more discernible interface
between layers, which consequently provides a more straightforward platform for fundamental investigations of material
properties and device physics. Charge-selective transporting (or
blocking) layers are oen employed in planar devices, just as in
the mesoscopic solar cells. This stacked-layer architecture is
also widely encountered in thin-lm solar cells and, to a certain
extent, in organic solar cells.9,10
In previous work performed by Ball et al., planar devices
based on the mixed halide CH3NH3PbI3xClx perovskite only
yielded a PCE of 4.9%, while the corresponding devices with the
incorporation of a mesoporous Al2O3 scaold were more than
twice as ecient.11 Nevertheless, the following revelations pertaining to the material properties of these hybrid halide materials: (1) the combination of both high dielectric constant and
low eective masses in a hybrid halide perovskite promotes the
formation of Wannier-Mott excitons, i.e., a type of exciton with
low binding energies and high likelihood to dissociate in bulk
even at room temperature;12 (2) electronhole diusion lengths
are long (exceeding 100 nm) and balanced;13,14 (3) high optical
absorptivity of organolead halide perovskites enables ecient
photon harvesting even with an ultrathin absorber layer,15,16
suggest that it is conceptually viable to obtain highly ecient
perovskite solar cells with a thin-lm planar architecture. Thus
despite the sluggish start, planar heterojunction perovskite
solar cells were soon found to catch up with the performance of
the mesoscopic ones. Device eciencies in excess of 15% even
as high as 19.3%, although most of these remain uncertied,
exhibit immense potential in these thin-lm planar heterojunction perovskite solar cells.5,1725
Organometal halide perovskite-based solar cells have presented to the solar cell community not only an exciting but also
8944 | J. Mater. Chem. A, 2015, 3, 89438969
Review
a challenging problem. The euphoria is mainly because of the

incredibly high eciencies achieved so far; those exceptional
results are not only limited to a few groups with special facilities, instead they have been reproduced by dierent research
groups all over the world. On the other hand, it is challenging
because of the incredibly fast pace at which the eciency is
evolving, which divert the attention away from fundamental
studies of these perovskite materials, their interfaces with the
other layers and the impact of the device architectures on the
morphologies of the lms. In this review, an attempt is made to
discuss various strategies to control the morphology of organometal halide perovskites. Together with this, the common
architectures employed in perovskite-based solar cells and how
they dictate the choice of interlayers and device performances
are also presented. Other review papers on similar topics have
been published previously by Gamliel and Etgar and by He
et al.26,27 We distinguish our work in terms of the extent of the
scope covered as well as the insights provided on the relevant
studies to date.
2.
Morphological control
In order to obtain good performance in planar perovskite

devices, the most direct approach implemented by dierent
groups thus far is to increase the thickness of the organolead
halide perovskite layer to maximize photon harvesting.
However, this is not as straightforward as it seems because the
lm quality and hence the shunting pathways and the absorption quality of the device are related to its crystallization
behaviour. Therefore, one of the main challenges in this eld is
the fabrication of high quality lms with controlled
morphology, high surface coverage and minimum pinholes for
high performance, solution-processed thin-lm perovskite
devices. Factors such as charge dissociation eciency, charge
transport and diusion length of charge species are dependent
on the crystallinity of the lm.14,28 The crystallization behaviour
of these perovskite materials is in turn highly dependent on
factors such as deposition methods, composition, type of
surfaces (surface chemistry, degree of hydrophilicity, surface
structure, etc.), and solvents/additives used. Until recently,
humidity has been shown to be detrimental to the performance
of the cells28 because of the hygroscopic nature of CH3NH3+ but
it was found that a controlled humidity of about 30% was able
to help with the reconstruction during the lm formation
process by partially dissolving the reactant species. This gives
rise to a spectacular eciency of more than 19%5 which enables
these cells to compete with conventional thin-lm solar cells.
In the case of mesoscopic solar cells, mesopores, regardless
of whether they are TiO2, Al2O3 or NiO, provide a physical
constraint on the crystal dimensions. This resulted in the
possibility of achieving good lm quality even with a reasonably
thick lm. This is the reason for the excellent device performance obtainable with this architecture.29 One important point
to note in the mesoscopic devices, is that a perovskite overlayer
is still necessary for good device performance.11 The crystals
within the perovskite overlayer have a much larger grain size
(1001000 nm), which is necessary for better charge transport.
View Article Online
Review
Ball et al. demonstrated clearly that the crystal size of the mixed
halide organolead perovskite CH3NH3PbI3xClx was reduced to
less than 100 nm in mesoporous Al2O3 compared to a planar
system in which the crystals can be more than 500 nm in size.
This results in an increase in the JSC for the planar device as
there is less charge trapping and recombination at grain
boundaries (VOC and FF are better with the MSSCs because of
the better lm formation and hence reduced shunting paths).11
This is contradictory to previous work published by Lee et al.
from the same research group claiming that the optimized
device is one without a capping layer.7 The reason for this
discrepancy is most probably because of the dierent processing conditions for Al2O3. In the previous paper by Lee et al., the
sintering temperature of Al2O3 was 500 C but in the later paper
by Ball et al., the temperature was lowered to 150 C. The crystallization behaviour is most likely dierent due to the dierence in the surface energy of the Al2O3 scaolds annealed at
dierent temperatures.
Due to the sensitivity of perovskite-based solar cells to their
lm morphology, controlled crystallization is an essential
consideration in the fabrication of these cells. Crystallization
is a complex process that involves two main steps nucleation
and growth. Nucleation from a continuous phase can occur
homogeneously or heterogeneously. The classical theory arises
from the work by Gibbs, Volmer and others.30 Homogeneous
nucleation is a matter of considering the overall excess free
energy. On the other hand, heterogeneous nucleation, which
occurs on foreign nuclei or surfaces, imposes a consideration
of surface or interface contact energy due to the wetting on
foreign surfaces. Therefore factors aecting surface properties, such as chemistry of the surfaces types of materials
(TiO2 vs. Al2O3 or PEDOT:PSS), thermal treatment on surfaces,
surface morphology (mesoporous vs. compact layer) become
very important in addressing this type of nucleation. Nucleation barrier is much lower for heterogeneous nucleation
compared to homogeneous nucleation because of the shape of
the nucleus. The free energy needed for heterogeneous
nucleation (DGheterogeneous) is related to that needed for
homogeneous nucleation (DGhomogeneous) by the following
expression and the terms are dened in the schematic shown
in Fig. 1:
DGheterogeneous DGhomogeneous f(q)
where
f q
2 3 cos q cos3 q
:
4
Simplied schematic of heterogeneous nucleation on solid

surfaces.
Fig. 1
This essentially means that the energy barrier for heterogeneous nucleation on solid/liquid interfaces is lower than that
for homogeneous nucleation. Better wetting (smaller contact
angle, q) will lower the nucleation barrier. It is consequently
expected that nucleation and hence growth of perovskite crystals will occur more easily when the solution is deposited on
substrates with larger surface area or more hydrophilicity since
perovskite materials are generally dissolved in more polar
solvents such as g-butyrolactone (GBL), N,N-dimethylformamide (DMF) and dimethylsulfoxide (DMSO). On top of this, in
the case of mesoporous scaolds, regardless if the scaold
participates in the electron transport, the curvature is towards
the liquid, the contact angle will be further reduced, hence
further reducing the energy barrier. The contradictory results
presented by Lee et al. and Ball et al. can therefore be explained
by the dierences in the crystallization process due to the
variation in surface properties.7,11 Dehydroxylation of the Al2O3
surface occurs upon heating of Al2O3 to above 500 C rendering
the surface less hydrophilic.31 This increases the barrier for
nucleation. Therefore, one of the criteria for uniform
morphology is to encourage heterogeneous nucleation to occur
and this can be done with the use of a mesoporous scaold
(large surface area), making the surface hydrophilic (reduced
dehydroxylation of the oxide lm).
For both mesoscopic and planar architectures, a variety of
methods have been demonstrated to control the growth kinetics
of the perovskite lm. Factors such as solution concentration,
precursor ratio, solvent, deposition temperature and lm
substrate interfacial interaction are found to aect the lm
formation.32 As mentioned previously, pinholes and incomplete
surface coverage of a perovskite lm in a device may result in
low-resistance shunting pathways (via direct contact of n-type
electron-transporting layer and p-type hole-transporting layer)
and contribute to the loss in photon absorption; both are
detrimental to the device performance. Therefore, eective ways
to manipulate nucleation and growth of perovskite crystals in
solution-processed systems, particularly in planar systems, in
order to achieve optimum lm morphology are highly sought
aer. The following are a few strategies that have been adopted
to control the morphology of perovskite lms.
2.1
Deposition methods
The morphology of an organometal halide perovskite lm, which

essentially comprises two main precursor components, depends
on the deposition route (solution-based or vacuum-based). The
precursors can either be simultaneously or independently
deposited. The combination of these permutations generates a
range of thin-lm deposition techniques for hybrid perovskite
materials. In general, without any additional connement, or any
modication to the composition or alteration to the crystallization process, single deposition produces a more uncontrollable
crystallization resulting in a variety of morphologies and hence a
spectrum of device performances. In a single deposition
approach, a mixture of metal halide and organohalide of certain
ratio is typically dissolved in a common solvent and spin-coated
on a substrate.8 This approach is the most cost eective and
J. Mater. Chem. A, 2015, 3, 89438969 | 8945
View Article Online
straightforward to implement. Nevertheless, eciencies of

planar devices prepared using this method are usually limited,
owing to poor lm formation (incomplete surface coverage with
pinholes) induced by extensive crystallization caused by solvent
evaporation and strong ionic interaction between the metal
cations and the halides.33,34 The choice of eective solvent
capable of simultaneously dissolving both precursors is also
limited. In our previous work, we managed to obtain PCE as high
as 5.2% with an inverted PEDOT:PSS/CH3NH3PbI3 (50 nm)/
PC61BM by employing a one-step solution approach.15 However
obtaining thicker but smooth and uniform lms by spin-coating
solutions of higher concentrations was a challenge.
The sequential deposition method oers the possibility of a
more controlled crystallization of perovskite.28,35 Metal halide
(e.g., PbI2) is rst introduced in the mesoporous TiO2 scaold; at
a relatively high concentration, PbI2 in the form of hexagonal 2H
polytype is constrained by the pores (22 nm). Upon exposure to
the organohalide (e.g., CH3NH3I) precursor solution, the
conversion to perovskite occurs immediately and is complete
within seconds. On the other hand, the conversion takes longer
on a planar substrate, i.e., incomplete conversion was observed
even aer 45 min. In a planar system, MAI needs to diuse into
the PbI2 lm prior to conversion. Perovskite crystallization
becomes kinetically more controllable because nucleation and
reaction rate between both precursors can be independently
manipulated. An additional heat treatment is performed to
remove residual solvent and to promote a more complete
conversion. This technique enables a highly ecient mesoscopic
perovskite solar cell device (PCE 15%) with improved reproducibility. We have also implemented this strategy for inverted
planar devices, resulting in improved eciency from 5.2% to
7.4%.15 The success of sequential deposition was also observed by
Hu et al., who employed this method in fabricating NiO-based
inverted planar perovskite solar cells.36 They claimed that the
metal halide lm had to be rst heated before being dipped into
a warm (70 C) organohalide solution. The sequentially deposited
CH3NH3I lm exhibited large crystals on the length scale of tens
of nanometers. The other variation of the two-step solution-based
deposition was demonstrated by Xiao et al.19 where a bilayer of
each component was deposited on top of each other and heated
to encourage the interdiusion of both precursor materials. The
degree of conversion depends on the amount of precursors
available (PbI2/CH3NH3I thickness ratio in the bilayer lm),
temperature and duration of the heat treatment. The optimized
device showed an eciency of 15.4% with high reproducibility
and no obvious hysteresis. The method was further modied by
Chiang et al. who spin-coated 4 times more slowly and found the
resulting lm required no heating.22 The slower spin-coating
prevented rapid solvent removal, which in turn assisted the
inltration of CH3NH3I molecules into the PbI2 lm. They
successfully obtained a planar device with a remarkable conversion eciency of 16.3% with no hysteresis observed.
Vacuum-based deposition oers potential solutions to the
limitations imposed by solution processing. Liu et al. used a
dual-source vapor deposition technique to prepare a mixed
halide CH3NH3PbI3xClx perovskite thin lm.17 The lm had
extremely high uniformity over a range of length scales on
8946 | J. Mater. Chem. A, 2015, 3, 89438969
Review
compact TiO2 with crystalline features on the length scale of

hundreds of nanometers. As a result, the vapor-deposited
sample yielded a high PCE of 15.4%, while the solution-processed equivalent only gave a PCE of 8.6%. However, in one-step
co-deposition it is dicult to simultaneously control the
deposition rates of both precursors, potentially leading to an
undesirable non-stoichiometric lm. A sequential vapor deposition method can be applied, as shown by Hu et al.37 A bilayer
lm is rst prepared from sequential deposition of PbI2 and
CH3NH3I, both of which should partially react in situ during the
vacuum-deposition of CH3NH3I, followed by post-deposition
thermal annealing to complete the perovskite conversion
process as well as to enhance lm texturing. The smooth polycrystalline perovskite lm with a grain size of 500 nm
demonstrated an eciency of 5.4% when used in devices with a
simple architecture with no interlayers (ITO/CH3NH3I/C60/Ag).
Chen et al. implemented in situ heating during the vapor
deposition of CH3NH3I to improve perovskite conversion and
crystallization, resulting in an ITO/PEDOT:PSS/CH3NH3PbI3xClx/C60/Bphen/Ca/Ag device with a PCE of 15.4%.38
Chen et al. presented a hybrid approach, combining both
solution- and vacuum-based approaches by using a vaporassisted solution process (VASP).39 Planar devices prepared
using this approach demonstrate a promising eciency of
12.1%. The PbI2 lm is rst deposited by spin-coating, followed
by an exposure to a CH3NH3I-vapor-rich environment at
elevated temperature. The CH3NH3I vapor is generated by
heating the precursor powder in a closed container. Akin to the
other two-step approaches discussed before, VASP takes
advantage of a more kinetically controlled CH3NH3I diusion
into a PbI2 framework to generate a more thermodynamically
stable compact perovskite lm with well-dened grains.
In brief, a two-step deposition method (whether solutionbased or vacuum-based or combination of both), despite its
intricacy, allows for independent control in the deposition
conditions of each precursor and hence their reaction thereaer. This eventually enables a better control over the perovskite lm formation process.
2.2
Solvent engineering
The solvent engineering approach demonstrated by Jeon et al. is

an attractive approach to circumvent the limitations of a onestep spin-coating process, which tends to yield an inhomogeneous perovskite lm hampering its reproducibility.29 The
perovskite lm is deposited from a precursor solution dissolved
in a mixture of dimethylsulfoxide (DMSO) and g-butyrolactone
(GBL) (DSMO : GBL 3 : 7 v/v), immediately followed by a
toluene drip while the substrate is spinning (Fig. 2). DMSO is
used due to its strong coordination with PbI2 to form a lead
halidesolvent complex, while GBL with higher boiling point
behaves purely as a solvent. A mixed-solvent approach alone has
also been shown by Kim et al., who used a combination of N,Ndimethylformamide (DMF) and GBL (DMF : GBL 97 : 3 v/v), to
prepare perovskite lms with improved morphology.40 Toluene
was chosen due to its miscibility with both DMSO and GBL yet it
does not dissolve the precursor materials. The toluene drip,
View Article Online
Review
Fig. 2 Schematics of the solvent engineering procedure to prepare the perovskite thin lm. Reproduced with permission from ref. 29. Copyright
Nature Publishing Group.
therefore, removes excess DMSO solvent and this encourages

supersaturation in the cast lm and hence fast nucleation.
PbI2DMSO impedes the rapid reaction between perovskite
precursors, leaving a highly uniform CH3NH3IPbI2DMSO
complex lm. This complex lm can be converted into a crystalline perovskite when DMSO escapes from the lm during
annealing. These intermediate complex species essentially
moderate the rate of formation of the nal perovskite. A planar
device with the conguration of FTO/TiO2/CH3NH3Pb(I1xBrx)3/
PTAA/Au exhibited a PCE of 9.1% under forward scan (shortcircuit to open-circuit) and 14.4% under reverse scan (opencircuit to short-circuit), which indicates a large hysteresis. By
contrast, the corresponding mesoscopic mixed halide perovskite device prepared on a 200 nm thick mesoporous TiO2
scaold showed a PCE of 15.8% and 15.9%, for forward and
reverse scans, respectively.
Reporting independently, Wu et al. showed that the use of
DMSO solvent itself was sucient to yield highly uniform PbI2
lms, which were sequentially immersed into the organohalide
solution to convert to perovskites.41 Indeed, the use of DMSO
could simultaneously solve two main issues associated with a
sequential deposition technique: (1) incomplete conversion of a
thick PbI2 lm to CH3NH3PbI3 especially in the absence of a
mesoporous scaold and (2) inconsistent degree of conversion
resulting in polydisperse crystal sizes. Both issues originate
from the lack of control in PbI2 crystallization leading to varying
crystal sizes and poor lm uniformity. Since it is a diusionlimited process, the conversion is largely dependent on
CH3NH3I penetration into the PbI2 lm. By using a more
strongly coordinating solvent such as dimethylsulfoxide
(DMSO), they were able to manipulate the rate of PbI2 crystallization via formation of PbI2(DMSO)2 complexes. The sequentially deposited CH3NH3I lm exhibited a smoother surface
with smaller crystals compared to the lm prepared with DMF,
a more weakly coordinating solvent with Pb2+. Consequently,
this improvement in the control of the crystallization process is
translated to higher reproducibility in the fabrication of DMSObased perovskite devices.
evaporation rate, which is in turn dependent on its boiling point

and the spin-coating conditions. Huang et al. introduced an
additional step of owing Ar gas over a semi-wet perovskite lm
during the spin-coating process to accelerate solvent drying.24
This gas-assisted step induces supersaturation that results in
the formation of a large number of nuclei in the perovskite
solution. Post-deposition annealing further promotes the
crystal growth, yielding lms with densely packed single crystalline grains. The planar device constructed using this
approach gave a PCE of up to 14% with good reproducibility,
while the conventionally spin-coated lm resulted in a much
lower PCE of 4.6%.
Other than using gas-assisted drying, rapid solvent removal
can also be achieved by using a non-dissolving solvent such as
chlorobenzene (CB). This solvent has to have good miscibility
with the host solvent but poor solubility for the perovskite
precursors. Xiao et al. implemented such one-step, solventbased methods to achieve highly uniform CH3NH3PbI3 lms.20
The planar devices prepared using this approach gave a high
PCE of 13.9%. The CB drip removes any excess DMF and rapidly
reduces the precursor solubility, resulting in supersaturation
conditions very rapidly and this drives crystal nucleation and
growth process. This approach greatly resembles the solvent
method (DMSO/GBL/toluene) developed by Jeon et al.29 Despite
the similarity in protocols, the results are very dierent. The
presence of DMSO that coordinates strongly with lead halide
prevents immediate crystallization upon toluene drip, as
observed from the lack of color change in the lm. On the other
hand, CB drip instantly promotes darkening, which suggests
instantaneous perovskite formation. In both cases, good lm
quality and hence good device performance can be achieved
and this may suggest that the coordinating solvent is less critical for uniform lms when fast nucleation has taken place. By
controlling the degree of supersaturation in the deposited
perovskite solution, either using a gas-assisted or solventassisted approach, it is possible to have a better control over the
crystallization process during the lm formation.
2.4
2.3
Rapid solvent removal
The unconstrained crystallization of a perovskite during deposition can be attributed to a certain extent to the slow solvent
Chloride introduction
Unlike their mesoscopic counterparts, planar perovskite

devices, based on organolead mixed halide CH3NH3PbI3xClx,
have enjoyed more success compared to the pure triiodide
J. Mater. Chem. A, 2015, 3, 89438969 | 8947
View Article Online
system.5 One possible reason to explain this observation is the

modulation of crystallization with the presence of chloride ions
and this has a more signicant eect on the planar system than
a mesoscopic system. Hence doping with chloride ions, from
either metal chloride (e.g., PbCl2) or organochloride (CH3NH3Cl), is another eective method to improve the morphology of
perovskite lms.7,42 The modied morphology in the mixed
halide perovskite may be the reason for its much longer electronhole diusion length, exceeding 1 mm, and roughly 10
times higher than that of its triiodide counterpart.13 The mixed
halide perovskite lm is typically prepared with a precursor
(PbCl2 : CH3NH3I) ratio of 1 : 3. Post-deposition heat treatment
induces chloride loss in the form of gaseous CH3NH3Cl with
higher volatility than CH3NH3I, resulting in the formation of
CH3NH3PbI3. Therefore, these Cl ions can facilitate the
removal of excess CH3NH3+ ions even at low annealing
temperatures.43 It has been reported that there is only a negligible amount of chloride phase in the nal perovskite lm. In
addition, the characteristic X-ray diraction (XRD) peaks of the
Cl-doped CH3NH3PbI3 are identical to those without
doping.4345
Zhao and Zhu demonstrated the preparation of CH3NH3PbI3
perovskite thin lms by incorporating CH3NH3Cl into a typical
precursor mixture of PbI2 and CH3NH3I.42 The presence of
CH3NH3Cl altered the kinetics of the crystallization process
(formation of PbCl2), which signicantly improved the perovskite lm morphology as observed from the disappearance of
sub-micron islands. The mixed halide lm also had improved
optical absorption. As a result, the planar devices exhibited an
improved PCE from 2% to 12% upon CH3NH3Cl incorporation.
Chloride ions can also be introduced by immersing a PbI2 lm
into a mixed solution of CH3NH3I and CH3NH3Cl.23 The chloride doping results in: (1) smooth morphology with distinct
crystals; (2) enhanced lifetime of the photoexcited species; (3)
improved light absorption, and (4) decreased series resistance
of the solar cell device; all of which contribute to a high PCE of
15%. Inorganic salt such as NH4Cl could also be employed as a
chloride source for organolead iodide perovskites, as shown by
Zuo and Ding.46 In comparison to CH3NH3Cl, NH4Cl resulted in
smoother lms with better crystallinity and improved optical
absorption. The inverted planar devices based on ITO/
PEDOT:PSS/CH3NH3PbI3/PC61BM/Al showed a much higher
PCE of 9.9% with NH4Cl additive, as compared to 8.1% (with
CH3NH3Cl additive) and 0.1% (without any additive).
The transient presence of chloride prompts the formation of
the chlorideiodide intermediate phase that alters the crystallization kinetics. These intermediate species possess distinct
crystalline features that can be observed at the early stages of
thermal annealing by looking at the X-ray diraction patterns.47
In an attempt to explain the possible origin of the intermediate
mixed halide phase, Tidhar et al. hypothesized that in the
mixture of PbCl2 and CH3NH3I, PbCl2 particles could act as
heterogeneous nucleation sites for the formation of perovskites.48 We believe this conjecture can be extended to the other
chloride-doped systems owing to the possibility of ionic
exchange between the dierent halide sources in solution.
Nevertheless, Williams et al. demonstrated that there were
8948 | J. Mater. Chem. A, 2015, 3, 89438969
Review
subtle variations in the kinetics of the complex ion formation

and aggregation in the solution depending on how the chloride
is introduced.49 The chemical equilibria in the precursor solution, or the lack thereof, might aect the crystallite formation
and transformation in the cast lm. They further proposed that
the chloride-rich intermediate species acted as templates facilitating topotactic self-assembly of the precursor materials into
perovskite crystallites, which undermines the eect of substrate
interfacial energy predominantly observed in pure iodide
systems, before being lost via sublimation or degradation in the
form of organochloride compounds during the annealing
process (Fig. 3).
The versatility of chloride incorporation can also be
extended to mixed iodidebromide organolead perovskites.44,45
An evolution in morphology from one-dimensional (1D)-like
crystals (e.g., bers or nanobelts) into a more uniform crystalline lm has been observed upon Cl inclusion. Similar to the
pure iodide system, the chloride entities are proposed to
promote the formation of trihalide intermediate species
comprising all the halide precursors, which subsequently alter
the kinetics of perovskite transformation.
Although chloride ions have been known to have a profound
eect on perovskite morphology, their other functions as well as
the verication of their perpetual presence in thin lms remain
disputable. There are, however, a few phenomena that can be
explained by assuming the actual presence of chlorides. By
performing density functional calculations, Du demonstrated
that the high carrier diusion length in the mixed chloride
iodide lm was attributed to the signicant increase in the
formation energy of the interstitial defects, in turn caused by
the reduced lattice constant due to the smaller Cl atoms.50 On
the other hand, Mosconi et al. discovered by rst principle
electronic structure calculations that interfacial Cl atoms at the
TiO2 interface might lead to a stronger electronic coupling
Fig. 3 (a) Schematics of morphological evolution in the CH3NH3PbI3xClx lm: nucleation during precursor solution deposition (top),
phase evolution and growth during annealing (middle), and nal
morphology (bottom); (bd) representative images of regions with
dierent morphological constituents in the CH3NH3PbI3xClx lm; the
scale bars are 5 mm. Reproduced with permission from ref. 49.
Copyright 2014 American Chemical Society.
View Article Online
Review
between mixed-halide perovskite and TiO2, enhancing electron

extraction between these materials.51
2.5
Precursor composition
Wang et al. observed a strong dependence of CH3NH3PbI3

morphology on the stoichiometry ratio of its corresponding
precursors.52 Deposited under the same conditions with lowtemperature solution processing on at PEDOT:PSS-coated
substrates, the equimolar solution (PbI2 : CH3NH3I 1 : 1)
demonstrates a rough morphology with microber formation.
Reducing the PbI2 content in the precursor solution, i.e.,
decreasing the precursor ratio, was observed to reduce the
formation of the microbers, gradually followed by the formation of a more compact lm with fewer pinholes. Furthermore,
it was also found that the precursor compositions in the spincoated lms were dierent from that in the solutions, which
could be due to the varying interaction strength between the
precursors and the substrate. The optimum precursor ratio was
also found to be dierent for lms with dierent thicknesses.
Another study on precursor composition was performed by
Yu et al. on both mixed halide (I/Cl) and pure iodide perovskites.43 A similar trend in the morphological evolution was
observed with the reduction of PbI2/CH3NH3I precursor ratio. It
was, however, hypothesized that the organohalide-rich environment in lms with low precursor ratios might have introduced certain intermediate phases, resulting in a retarded
crystallization rate.
2.6
Solvent additive
The crystallization rate of perovskites can also be manipulated

by incorporating additive molecules into the precursor solution.
Two important criteria to consider in the selection of solvent
additives are: (1) higher boiling point than that of the host
solvent and (2) preferential interaction with the precursor
materials. One strategy to modulate the interaction with
precursors is by taking advantage of the chelating properties of
molecules such as bidentate halides to metal ions such as Pb2+.
It is therefore possible to alter the crystallization kinetics and
hence morphology of perovskites. Liang et al. employed 1,8diiodooctane (DIO) due to its high likelihood to coordinate with
Pb2+ (from PbCl2) during crystal growth as well as its higher
boiling point (Tb 332 C) as compared to the host solvent DMF
(Tb 153 C).32 As a result, the interaction between DIO and
Pb2+ becomes more prominent as most of the host solvent is
removed during spin-coating. The chelation of DIO with PbCl2
might have induced the formation of a more thermodynamically stable intermediate structure, which leads to the modication of crystallization rate. The DIO-treated lm exhibits
smooth morphology with improved surface coverage.
2.7
Thermal annealing and ash annealing
Heat treatment performed on perovskite lms typically serves a

few purposes: (1) to remove residual solvent from solution
processing; (2) to assist perovskite formation from its precursors, and (3) to enhance crystallization and grain growth.
Depending on the conditions employed during the thermal
annealing process (e.g., temperature, duration, ambient atmosphere) as well as the attributes of the perovskite lms exposed
to the heat treatment (e.g., lm thickness, lm composition),
distinct morphologies could be obtained. A mixed halide
perovskite typically has to be annealed longer to achieve full
conversion compared to the single halide system. Hybrid
perovskite lms that survive thermal annealing can be considered to be more thermally stable at room temperature. Eperon
et al. were able to optimize the morphology of the CH3NH3PbI3xClx perovskite lm deposited on planar substrates by
ne-tuning the thermal annealing conditions.28 Both prolonged
heat treatment and annealing at higher temperature typically
caused coarsening of the perovskite crystals with decrease in
surface coverage while the use of thicker lms improved surface
coverage upon annealing. Based on these results, it may be
sensible to use a thick CH3NH3PbI3xClx lm and anneal it at
lower temperatures to improve surface coverage. Dualeh et al.
found out that a minimum temperature of 80 C was required to
convert the PbCl2 and CH3NH3I mixture into the CH3NH3PbI3xClx perovskite.53 Similar to the observation obtained by
Eperon et al., annealing at higher temperatures induced the
island formation, which was accompanied by increased
content of PbI2 phase via CH3NH3I loss, resulting in a decrease
in device performance.
The conventional thermal annealing treatment of perovskite
can be coupled with ash annealing, i.e., rapid ramping to a
higher temperature followed by brief heating; the latter step
modies the perovskite morphology.54,55 Saliba et al. showed
that ash annealing was more eective for planar architecture, while the mesoscopic device suered from reduced surface
coverage.54 Therefore, it is crucial to consider the dierent types
of device congurations when devising heat-treatment methodologies. Expectedly, the ash annealed planar device
demonstrated an improved eciency.
2.8
Solvent annealing
Solvent annealing has been demonstrated to be an attractive

strategy to optimize the morphology of organic solar cells.56
Essentially, a semiconductor lm is placed in an environment
saturated with solvent vapor. The rst demonstration of solvent
annealing for organolead perovskite lms was given by Xiao
et al., who annealed CH3NH3PbI3 in the presence of DMF
vapor.25 By employing solvent annealing, the perovskite crystals
can grow more extensively because the DMF vapor, in which
both PbI2 and CH3NH3I precursors are highly soluble, can
facilitate the diusion and reorganization of the precursor
molecules and ions. Interestingly, it was also found that the
crystal growth was dependent on the lm thickness, i.e., thicker
lms (more materials present) could produce larger grains of
about 1 mm (Fig. 4). The inverted planar devices prepared with
solvent-annealed CH3NH3PbI3 did not exhibit a strong dependency of device performance on lm thickness. Even the
thickest perovskite lm prepared (1 mm thick) was able to yield
PCE approaching 15%, which was comparable to those of the
thinner lms.
J. Mater. Chem. A, 2015, 3, 89438969 | 8949
View Article Online
Review
formation of lms with good uniformity and thereaer, growth

should be controlled. Nucleation can be encouraged by
achieving the supersaturation conditions quickly either
through rapid solvent removal or through the fast formation of
some intermediate nuclei. Crystallization of the lms can be
promoted using post-deposition annealing techniques to obtain
crystals of optimal size.
3.
Device architecture
The device performance of perovskite solar cells is highly

dependent on their architecture which will in turn dictate the
choice of materials, the deposition methods for the material
and naturally, the compatibility between the dierent components in the device. Two major embodiments of perovskite solar
cells have been developed so far, i.e., mesoscopic and planar
structures. In the mesoscopic architecture, the perovskite can
either be introduced as a thin layer that will just adequately
cover the oxide scaold with the pores in the scaold inltrated
with charge transporting material (Fig. 5a), or the perovskite
can form an overlayer on top of the completely inltrated oxide
scaold (Fig. 5b). A more straightforward planar architecture
can exist in either conventional (Fig. 5c) or inverted (Fig. 5d)
conguration, depending on the direction of electric current.
3.1
SEM images of the thermally annealed (TA) and solventannealed (SA) perovskite lms with thicknesses of (a) 250 nm, (b) 430
nm and (c) 1015 nm; the scale bars in the SEM micrographs are 2 mm.
Reproduced with permission from ref. 25. Copyright 2014 Wiley-VCH.
Fig. 4
2.9
Humidity eect
Ambient conditions, such as the amount of solvent vapor or

humidity, may inuence the crystallization of hybrid perovskites. In particular, humidity is an important factor to consider
due to the ease of organometal halide perovskites to degrade in
a highly humid environment. Zhou et al. demonstrated that the
mixed halide CH3NH3PbI3xClx perovskite underwent a
dierent conversion process when annealed under controlled
humidity (30% RH).5 CH3NH3PbCl3 phase could be clearly
observed in the lm annealed with moisture involvement,
which could not be found in lms made under dry air. The
presence of a small amount of moisture alters the transformation route by partially dissolving the reacting entities and
enhancing mass transport in the lm. This so-called enhanced
reconstruction strategy resulted in improved PCE from 13.5%
to 16.4% for the TiO2/perovskite/spiro-OMeTAD device. The
same strategy has also been successfully applied in solar cells
with dierent architecture (PEDOT:PSS/perovskite/PC61BM),
boosting PCE from 12.3% to 15.4%.57
Perovskite lms with excellent uniformity and large crystal
size can be obtained through careful control of the nucleation
and growth process. Based on the discussion so far, the general
conclusion is that nucleation should be encouraged for the
8950 | J. Mater. Chem. A, 2015, 3, 89438969
Mesoscopic architecture
Since the initial evolution of perovskite solar cells was based on

a mesoporous (mp) structure, there are already several review
papers on mesoscopic perovskite solar cells.5867 Thus, in this
section, we will only focus our discussions on metal oxide
scaolds and hole-transporting materials (HTMs) for mesoscopic perovskite solar cells to avoid repetition of the existing
accounts of this type of solar cell. As presented previously in the
Introduction section, by considering the functions of the metal
oxide scaold, all the mesoscopic perovskite solar cells
demonstrated so far can be classied into the following two
groups.
3.1.1 Mesoscopic perovskite solar cells with active scaffolds. Of the active scaolds for mesoscopic perovskite solar
cells studied so far, TiO2 nanoparticles have been the most
commonly employed materials. It is still under debate whether
the presence of a scaold is a prerequisite to obtain perovskite
solar cells with enhanced device performance. However, many
publications suggest that the eciency of mesoscopic perovskite solar cells is strongly related to the thickness of the TiO2
scaold layer. By investigating CH3NH3PbI3 solar cells with a
mp-TiO2 layer ranging from 0.6 mm to 1.5 mm, Kim et al. found
that quite unlike solid state DSSCs which required a thick TiO2
scaold (3 mm) to achieve sucient absorption, hybrid lead
iodide perovskite sensitized solar cells only need a submicron
thick mesoporous TiO2 layer to achieve good performance.8 The
best performance of over 9% was from the device with at least
600 nm thick mp-TiO2 layer. Leijitens et al. attributed the
dependence of the device performance on the thin layer of
mesoporous TiO2 layer to a perovskite pore lling eect
(Fig. 6).68 They found that the pore lling fractions were heavily
View Article Online
Review
Fig. 5 Schematic diagram of mesoscopic heterojunction solar cells (a) no perovskite overlayer and (b) with perovskite overlayer; and planar
heterojunction solar cells with (c) conventional nip and (d) inverted pin congurations.
dependent on the TiO2 scaold thickness. When certain

concentrations of perovskite precursor solutions were used for
deposition, there was always an accompanying required thickness of mp-TiO2 layer so that perovskite could form continuous
lms onto the mesoporous TiO2 rather than discrete small
nanoparticles. In general, a perovskite precursor solution with
higher concentration can only completely ll the pores in a
thinner mp-TiO2 lm. Based on their calculations, for high

concentration of perovskite precursor solution (40 wt%), thin
mp-TiO2 (300 nm) is needed to ensure high device performance. A completely perovskite-lled TiO2 architecture can lead
to high device performance because there will be high electron
density in the TiO2, improving the charge transport rates and
collection eciency. In addition, the recombination between
Fig. 6 The transport lifetimes of devices with various pore lling fractions as a function of light intensity obtained from small perturbation
photocurrent decays, and the proposed recombination mechanism for solar cells with incomplete and complete perovskite pore lling.
Reproduced with permission from ref. 68. Copyright 2014 American Chemical Society.
J. Mater. Chem. A, 2015, 3, 89438969 | 8951
View Article Online
TiO2 and HTM could be minimized by the presence of the

perovskite over all the TiO2 surface.
It should be noted that besides TiO2 nanoparticles, TiO2
nanocrystals in other shapes (e.g., nanorods, nanowires, nanobers, nanotubes and nanocones) and similar n-type materials
such as ZnO have been applied to make the scaold for mesoscopic perovskite solar cells, which can also act as an ecient
electron collector.6975 In particular, solar cells employing singlecrystalline rutile TiO2 nanocones were observed to outperform
the nanorod-based devices due to a more ecient electron
transfer from perovskite to TiO2.75 Furthermore, active scaold
materials are not only limited to n-type semiconductors. Tian
et al. rst demonstrated a new type of ecient mesoscopic
perovskite solar cell based on mesoporous p-type metal oxide
NiO.76 It was found that the FTO/NiOx/mp-NiO/CH3NH3PbI3/
PC61BM/BCP/Al device (where PC61BM phenyl-C61-butyric acid
methyl ester) could achieve a PCE of 1.5%. Recently, Wang et al.
obtained a more impressive PCE of 9.51% with similar device
architecture (ITO/NiOx/mp-NiO/CH3NH3PbI3/PC61BM/BCP/Al,
where BCP bathocuproine).77 Transient absorption and photoluminescence studies conrmed that ecient charge transfer
occurred at the NiO/perovskite heterojunction. They further
improved the eciency of the device PCE to 11.6% by employing a low-temperature sputtered NiOx compact lm, instead of a
solution-processed one.78 The use of a p-type mesoporous layer
for hole transport and collection shown in these studies
increases the diversity of perovskite solar cells, and this is also a
good way to remove the unstable hole-conducting organic
materials oen used in mesoscopic perovskite solar cells.
Active scaolds can also be decorated with quantum dots
(QDs) such as lead sulde (PbS) to achieve co-sensitization with
an organolead halide perovskite that leads to a panchromatic
response from the visible to the near-infrared (NIR) region.79
Quantum dots not only improve the photon harvesting, but they
can also assist charge transport from perovskite to TiO2. Besides
semiconducting QDs, C60-SAM, graphene and its composites
have also been applied as an interfacial layer to improve charge
transfer between the perovskite and the active scaold.8082 The
use of the TiO2graphene composite, for instance, resulted in a
low-temperature processed electron collecting layer with a
perovskite solar cell demonstrating a remarkable eciency of
15.6%. Therefore, material selection and combination in
systems employing active scaolds are important considerations for obtaining highly ecient solar cells.
3.1.2 Mesoscopic perovskite solar cells with passive scaffolds. One of the main advantages of mesoscopic perovskite
solar cells with a passive scaold is their low-temperature
solution processability. When an Al2O3 passive scaold was
used in perovskite solar cells, high temperature (500 C) sintering was required to remove the organic binders inside the
thin lm, very similar to the preparation of mesoporous TiO2.
Later, Ball et al. introduced a low-temperature approach to
prepare an Al2O3 mesostructured scaold via solution deposition of Al2O3 nanoparticles without any binder, which were
subsequently sintered at 150 C to form the Al2O3 lm.11 They
demonstrated that the CH3NH3PbI3xClx perovskite solar cells
prepared on the low-temperature-processed Al2O3 scaold
8952 | J. Mater. Chem. A, 2015, 3, 89438969
Review
could achieve a PCE of up to 12.3%, with a VOC of 1.02 V, JSC of

18 mA cm2, and FF of 0.67. It was also the rst time that the
multifunctional roles of perovskite (light absorption, charge
generation, ambipolar electron and hole transport) in solar cells
were clearly shown, which was partially responsible for leading
to the development of scaold-less planar perovskite solar cells.
In subsequent work, Wojciechowski et al. demonstrated that it
is possible for all the layers in these mesoscopic perovskite solar
cells to be processed at low-temperature by reducing the processing temperature for a highly crystalline TiO2 compact layer
to less than 150 C.83 The optimized device showed a maximum
PCE of 15.9%. This renders mesoscopic perovskite solar cells
even more cost-competitive and compatible with exible
substrates. It should be mentioned that besides Al2O3, insulating ZrO2 nanoparticles have also been employed to construct
the mesoporous layer for CH3NH3PbI3 and the resulting solar
cells could give a PCE of 10.8%.84 Hwang et al. demonstrated the
use of SiO2 nanoparticles as a scaold layer. By employing
nanoparticles with an optimum size of about 50 nm, they could
obtain better perovskite inltration and hence achieve a PCE of
11.5% with a VOC of 1.05 V, which was higher than those of the
TiO2 nanoparticle devices.85
Similarly, the eect of the thickness of the passive scaold
on the device performance of mesoscopic perovskite solar cells
has been studied. Ball et al. have demonstrated that Jsc of FTO/
bl-TiO2/mp-Al2O3/CH3NH3PbI3xClx/HTM/Ag (where bl-TiO2
refers to the TiO2 blocking layer) can be tuned by varying the
thickness of the mesoporous Al2O3 layer (Fig. 7).11 JSC of these
perovskite solar cells increased slowly with the reduction in
Al2O3 thickness, reaching a maximum at a thickness of 80 nm
(16.9 mA cm2). The cross-section of these devices showed that
a capping layer of perovskite formed on the mesoporous Al2O3
layer when the scaold thickness was less than 400 nm. The
presence of the Al2O3 scaold can also act as a buer layer to
impede current leakage between the electrodes. In a very recent
study, Mei et al. did a detailed investigation on the role of the
perovskite capping layer in mesoscopic perovskite solar cells.67
It was found that the improved photocurrent when the capping
layer is present was to a large extent due to the thick perovskite
layer present in the system which enhanced the light absorption
and larger perovskite particles in the capping layer which
enhanced the light scattering.
In terms of the working mechanism, Kim et al. concluded
from their impedance spectroscopy measurements that perovskite solar cells on a passive scaold (ZrO2) behaved similarly to
a thin-lm solar cell, in which the light absorber also works
simultaneously as ambipolar electron and hole transporters, in
contrast to perovskite solar cells on an active scaold (TiO2) that
behaved halfway between a mesoscopic DSSC and a thin-lm
solar cell.86 The other feature of mesoscopic perovskite solar
cells fabricated with a passive scaold is the absence of photovoltage loss observed in their active-scaold counterparts. As
reported by Lee et al., comparable perovskite devices with TiO2
and Al2O3 scaolds exhibited a VOC of 0.8 V and 0.98 V,
respectively.7 The dierence in photovoltage is attributed to the
variation in chemical capacitance of the oxide, which is
essentially related to the charge-storing capacity caused by the
View Article Online
Review
Fig. 7 Cross-sectional SEM images of FTO/bl-TiO2/mp-Al2O3/perovskite/HTM/Ag solar cells with dierent thicknesses of the Al2O3 scaold,
and the dependence of device parameters on the scaold thickness.11
presence of sub-band gap states. Insulating Al2O3, inherently

lacking of surface and sub-band gap states, results in a significant decrease in chemical capacitance of the solar cell. These
states, which can also act as charge trapping sites, can inuence
the long-term stability of the device, which is discussed in
Section 4.
Recently, Edri et al. used the electron beam-induced current
(EBIC) technique to study the working mechanism of CH3NH3PbI3xClx solar cells.87 The appearance of two EBIC peaks at
the perovskite/electron acceptor (compact TiO2 layer) and
perovskite/hole acceptor (HTM) junctions regardless of the
presence of the insulating mesoporous Al2O3 layer indicates
that CH3NH3PbI3xClx solar cells have a pin characteristic
with an inhomogeneous built-in eld, which is independent of
the alumina scaold. This is direct evidence that the mesoporous Al2O3 layer in perovskite solar cells acts only as a scaffold. Besides, the higher photovoltage in mesoscopic perovskite
solar cells with an inert scaold is believed to arise from the
mutual contribution of both half junctions in series, each of
which could minimize the dark current and maximize the
photocurrent.
3.1.3 Hole transport materials (HTMs) for mesoscopic
perovskite solar cells. As discussed previously, spiro-OMeTAD is
the most commonly used hole HTM for perovskite solar cells,
mainly because it has been well studied as a HTM in OLEDs88
and solid state DSSCs.64,89 Solar cells based on hybrid lead
halide coupled with spiro-OMeTAD as a HTM have achieved a
PCE of more than 15%.17,35 However, the cost of spiro-OMeTAD
remains unfortunately high mainly because of its lengthy and
low yielding synthesis. The limitation in HTM options has
proved to be a major impediment to the growth and advancement of high eciency, cost-eective perovskite solar cells.
Hence it has become imperative to design and develop more

economical alternative HTMs to spiro-OMeTAD. In view of this,
much attention has been paid to the development of new HTMs
in recent years. Mesoscopic perovskite solar cells have so far
been used as the benchmark for many novel HTMs, and thus we
summarize the progress in this part.
Similar to solid-state DSSCs (ssDSSCs), HTMs for perovskite
solar cells should primarily meet the following requirements:
(1) good hole mobility; (2) a compatible HOMO energy level
relative to organolead halide perovskites for ecient hole
transfer; (3) good solubility and lm forming properties, and (4)
low cost. However, in perovskite solar cells, a HTM is not
required to ll in the pores, which are already lled by the
perovskite materials, hence eliminating the pore-lling issues
commonly encountered in ssDSSCs, which widens the choice of
materials. Three categories of HTMs have been investigated in
perovskite solar cells: inorganic, polymer and small molecular
HTMs. The chemical structures of the HTMs recently used in
perovskite solar cells and their corresponding HOMO levels are
respectively shown in Fig. 8 and 9.
The progress in inorganic HTMs, however, has been slow
mainly due to the limited choice of materials. Well known
inorganic HTMs such as CuI (PCE 6%),90 NiO (PCE 11.6%)78
and CuSCN (PCE 12.4%)91 have been investigated in mesoscopic perovskite solar cells. Though potentially cost-eective
and stable under ambient conditions, these HTMs have shown
inferior performance characteristics as compared to their
organic counterparts.
Several polymer HTMs including poly(triarylamine) (PTAA)
and poly(3-hexylthiophene-2,5-diyl) (P3HT), which generally
have better hole mobilities than small molecule HTMs and
good lm-forming properties, have been tested by Heo et al.;
J. Mater. Chem. A, 2015, 3, 89438969 | 8953
View Article Online
Review
Fig. 8 Structures of some of the recently reported HTMs for perovskite solar cells.
among them, PTAA demonstrated the highest eciency of up to

12% compared to 6.7% for P3HT.92 This was the rst major
breakthrough in the development of polymer HTMs, which
suered from pore-lling ability as reported for ssDSSCs. Later
the same group used a mixed-halide perovskite with PTAA and
obtained a certied PCE of 16.2% (JSC 19.6 mA cm2, VOC
1.11 V and FF 0.74); an even higher certied PCE of 17.9% has
also been recently reported.29,93 Chen et al. reported a P3HT/
MWNT (multi-walled carbon nanotube) composite as a HTM
which showed higher PCE (6.45%) than pristine P3HT (PCE
4.1%).94 MWNTs in a P3HT/MWNT composite act as ecient
nanostructured charge transport tunnels and induce crystallization of P3HT, hence signicantly enhancing the conductivity
of the composite. The ll factor of the hybrid solar cells was
greatly enhanced from 45% to 57%. Similarly, Habisreutinger
et al. employed P3HT-functionalized single-walled carbon
nanotubes (SWNTs) to substitute the traditionally used organic
HTMs, showing a high PCE of up to 15.3% with an average
eciency of 10%.95 Xiao et al. reported that polyaniline
(PANI), which acted as both sensitizer and HTM, gave a PCE of
7.34%.96 The PCE only reduced slightly (from 7.3% to 6.7%)
aer 1000 hours, indicating that the device has good long-term
stability.
Unlike inorganic HTMs, organic-based HTMs have exibility
in tuning the oxidization potentials and the surface properties.
Fig. 9
One of their attractive features is that high VOC can be achieved

with the combination of CH3NH3PbBr3 perovskite light
absorber and HTMs with deeper HOMO levels. Initial research
has been done by Cai et al. using a diketopyrrolopyrrole (DPP)carbazole based polymer HTM (PCBTDPP).97 The polymer HTM
has a HOMO level of 5.4 eV and demonstrated a VOC of 1.16 V.
An even higher VOC of 1.5 V was reported recently with 4,40 bis(N-carbazolyl)-1,10 -biphenyl (CBP).98 The CH3NH3PbBr3
perovskite typically lacks light absorption in longer wavelength
and is thus unfavorable to obtain a high eciency. However a
moderate PCE (6.7%) and a high voltage output (1.4 V) were
achieved by Ryu et al. with a triarylamine-based polymer HTM
(PIF8-TAA).99 Another interesting feature of polymer HTMs was
demonstrated by Kwon et al. with a hydrophobic DPP based
polymer (PDPPDBTE).100 This polymer has a deep HOMO level
of 5.4 eV with a good hole mobility of 103 cm2 V1 s1; hence,
the device showed a better PCE of 9.2% than 7.6% for spiroOMeTAD. The hydrophobicity of PDPPDBTE, veried by water
contact angle measurement, should contribute to the good
device stability that will be presented in Section 4.
As compared to the previous two types of HTMs, there are
more small molecular HTMs reported so far because of the
greater degree of freedom available for their design. In
designing small molecule HTMs, incorporating a triarylamine
component seems to be essential. The nitrogen center acts as a
Energy level diagram showing HOMO levels of various HTMs.
8954 | J. Mater. Chem. A, 2015, 3, 89438969
View Article Online
Review
doping site, which can be stabilized through conjugation with

neighboring aromatic rings. The non-planar conformation of
the triarylamine leads to longer intermolecular distances, which
may result in a low hole mobility (105 to 104 cm2 V1 s1), but
in return, makes the molecules less crystalline and more suitable as a HTM.
Several groups have developed novel molecules in which
triarylamine moieties were connected through an aromatic
linker. Jeon et al. synthesized a few pyrene-core arylamine
derivatives in which a pyrene core was substituted with various
numbers of N,N-di-p-methoxyphenylamine moieties.101 The best
cells fabricated using these pyrene derivatives (Py-C) as a HTM
showed a PCE of 12.4% under AM 1.5G illumination; one of the
highest PCE values reported with small molecular HTMs other
than spiro-OMeTAD, hence suggesting that very simple structures might still prove very useful as HTMs. Li et al. synthesized
a new HTM, having an ethylenedioxythiophene (EDOT) core
substituted with triphenyl amine moieties (H101), yielding a
PCE of up to 13.8%.102 The synthesis of this compound is much
simpler and more ecient as compared to that of spiro-OMeTAD. Krishnamoorthy et al. reported a swivel-cruciform 3,30 bithiophene-based HTM (KTM3) with a low lying highest
occupied molecular orbital (HOMO) level.103 This HTM, which
exhibited varying performance depending on the p-type
dopants used, demonstrated an optimized PCE of 11.3%.
Krishna et al. synthesized novel amorphous HTMs based on a
bulky triptycene core benecial to prevent crystallization.104 The
triptycene core is connected to diphenylamines via phenyl with/
without thienyl groups and the device based on one of the
derivatives (T103) showed a PCE of 12.4%. Jeon et al. reported a
PCE of 16.7% with a spiro-OMeTAD isomer with some of the
methoxy substituents in the ortho-positions.105 This is one of the
highest PCEs reported for any novel small molecule HTM. Choi
et al. have synthesized a fused quinolizino acridine based-HTM
(OMeTPA-FA) which produced a PCE of up to 13.6%.106 Do et al.
synthesized star-shaped HTMs based on an electron-decient
triazine unit which showed a PCE of up to 12.5%.107
Most of the abovementioned organic HTMs require additives
such as Li salt, tBP or cobalt complex to improve their
conductivity, hence enhancing the performance. Several groups
have reported perovskite solar cells without incorporating
additives into the HTM, which simplies the fabrication
procedures. Liu et al. reported the tetrathiafulvalene derivative
as a HTM (TTF-1).108 A PCE of over 11% was obtained without
the use of any dopants with this nitrogen-free HTM. In addition,
the lifetime of this cell under ambient conditions is 3 times
higher than the cells with spiro-OMeTAD, possibly a result of
this dopant free strategy. Qin et al. synthesized a donor
acceptor type quinolizino acridine-based HTM (fused-F) which
showed a PCE of up to 12.8%.109 This HTM has a low bandgap
and the LUMO level is close to that of the perovskite system,
hence it serves both the function of a light harvester as well as a
hole conductor.
So far various approaches adopted by dierent groups in
designing the HTM have seemed to give comparable results to
those with the standard HTM (spiro-OMeTAD). Eective
designing as well as optimizing thickness and additives are vital
factors to obtain high eciency. Future work needs to be done

to study the HTM/perovskite interface,110,111 which will shed
some light on the underlying device physics and will help in the
design of future molecules.
3.1.4 Hole-transporting material (HTM)-free mesoscopic
perovskite solar cells. Although various novel HTMs have shown
promising results, it is noteworthy to mention recent increasing
attention towards HTM-free perovskite solar cells. Excellent
semiconducting properties (i.e., long charge transport lifetime
and ambipolar nature)13,14 of organolead halide perovskite
materials enable the elimination of the hole transport layer;
holes can be extracted by a suitable counter electrode. The
simpler architecture in HTM-free devices leads to the reduction
of the processing cost, which is advantageous for practical
usage as they match with the low-cost solar cell applications and
eliminate the air-sensitive spiro-OMeTAD. Etgar et al. demonstrated the initial study employing anatase TiO2 nanosheets as
the mesoporous layer.112 The corresponding perovskite solar
cell (FTO/c-TiO2/mp-TiO2/CH3NH3PbI3/Au) displayed an eciency of 5.5% (JSC 16.1 mA cm2, VOC 0.63 V and FF 0.57).
It was revealed that in this type of solar cell, a depletion layer
extends to both CH3NH3PbI3 and TiO2 lms.113 The build-in
eld of the depletion layer assists in the charge separation and
suppresses the back reaction of electrons from the TiO2 to the
CH3NH3PbI3 lm. The device eciency also strongly correlates
with the width of the depletion layer, with the best performance
achieved upon depletion of half of the TiO2 lm.114 Further
optimization of the quality of the perovskite layer resulted in
eciencies of 811%.113118
Carbon electrodes can be used to replace noble metal
counter electrodes (e.g., Au or Ag) because of their earth abundance, large-scale printing processing, cost eectiveness and a
more suitable work function of 5.0 eV (5.1 eV for Au).119122 A
mesoporous carbon electrode was screen-printed on a double
layer of mp-TiO2 and ZrO2, followed by inltration with perovskites through the carbon layer; a PCE of 6.6% was achieved
with a device made with this electrode (FTO/c-TiO2/mp-TiO2/
ZrO2/CH3NH3PbI3/C).119 An improvement of PCE of up to 12.8%
(JSC 22.8 mA cm2, VOC 0.86 V and FF 0.66) was obtained
by incorporating 5-ammoniumvaleric acid (5-AVA) to prepare a
FTO/c-TiO2/mp-TiO2/ZrO2/(5-AVA)x(CH3NH3)1xPbI3
solar
cell.120 The 5-AVA cations act as the templates for the formation
of the perovskite crystals on the mesoporous scaold, resulting
in a more directed growth in the normal direction and better
connected crystals, hence lowering series resistance and
increasing electrical conductivity. Even when unsealed, these
HTM-free perovskite solar cells still demonstrate excellent
ambient stability due to the protection imparted by the carbon
electrode.
Various research studies have clearly demonstrated that
perovskite solar cells work as a pin device and thus a holetransporting layer is no longer a prerequisite for device operation.87,123 However, the device performance of HTM-free perovskite solar cells is still inferior to the state-of-the-art perovskite
cells because of the poor eciency of charge extraction and the
charge recombination at the interface; those cells typically show
lower VOC and FF. To improve charge extraction and to suppress
J. Mater. Chem. A, 2015, 3, 89438969 | 8955
View Article Online
Review
the undesirable recombination, strategies such as controlling

built-in eld,113,115 improving surface morphology,116,117 and
using insulating layers such as zirconium dioxide119,120 have
been demonstrated. This new device concept has greater
potential for practical applications and further improvement in
performance can be expected. A comparison of the resultant
device performance with dierent HTMs, including HTM-free
architecture, is shown in Table 1.
Table 1
3.2
Planar architecture
Although the embodiment of sub-150 C processing has been

recently reported for mesoscopic perovskite solar cells via the
utilization of nanoparticles,11,83,124 most literature reports still
employ mesoporous lms that require high-temperature sintering (500 C), particularly for the most ecient devices.
Therefore the lack of mesoporous metal-oxide scaold
contributes to the low-temperature processability of planar
heterojunction (PHJ) perovskite solar cells, inevitably imparting
A few representative HTMs and their corresponding device performances
Material
Typea
Architecture
HOMO/VB
level (eV)
VOC
(V)
JSC
(mA cm2)
FF
PCE
(%)
Reference
PCE (%)
SpiroOMeTAD
MS
mpTiO2/MAPbI3/HTM
5.22
0.99
20
0.73
15
MS
TiO2/mpAl2O3
/MAPbI3xClx/HTM
Y:TiO2/MAPbI3xClx/HTM
1.02
21.5
0.71
15.9
1.13
22.8
0.75
19.3
5.4
1.01
0.73
0.96
19.7
14.4
19.8
0.62
0.62
0.61
12.4
6.4
11.6
4.9
0.79
0.88
0.99
14.2
16.3
15.59
0.65
0.64
0.72
7.3
9.11
11.1
9.26
(PEDOT:PSS)
5.05.2
0.73
12.6
0.73
6.7
0.98
19.1
0.66
12.4
5.14
1.11
19.6
0.74
16.2
5.23
5.51
0.78
1.4
14.48
6.1
0.65
0.79
P
CuSCN
NiO
Graphene
oxide
P3HT
MS
P
MS
P
P
P
mpTiO2/MAPbI3/HTM
TiO2/MAPbI3xClx/HTM
mpNiO/MAPbI3/
PC61BM/BCP
HTM/MAPbI3xClx/PC61BM
HTM/MAPbI3/PC61BM
HTM/MAPbI3xClx/
PC61BM/ZnO
5.3
MS
mpTiO2/MAPbI3/HTM
TiOx/MAPbI3xClx/HTM
PTAA
MS
PANI
PIF8-TAA
MS
MS
mpTiO2/MAPb(I1xBrx)3/
HTM
mpTiO2/MAPbI3/HTM
mpTiO2/MAPbBr3/HTM
PEDOT:PSS
HTM/MAPbI3xClx/
PC61BM/PFN
4.95.2
1.05
20.3
0.80
17.1
PyC
MS
mpTiO2/MAPbI3/HTM
5.11
0.89
20.2
0.69
12.4
H101
MS
mpTiO2/MAPbI3/HTM
5.16
1.05
19.1
0.65
13.2
T103
MS
mpTiO2/MAPbI3/HTM
5.33
0.99
20.3
0.62
12.4
po-spiro
OMeTAD
OMeTPA-FA
MS
mpTiO2/MAPbI3/HTM
5.22
1.02
21.2
0.78
16.7
MS
mpTiO2/MAPbI3/HTM
5.14
0.97
21
0.67
13.6
TTF-1
MS
mpTiO2/MAPbI3/HTM
5.05
0.86
19.9
0.64
11
fusedF
MS
mpTiO2/MAPbI3/HTM
5.23
1.04
17.9
0.68
12.8
DR3TBDTT
TiO2/MAPbI3xClx/HTM
5.39
0.95
15.3
0.6
8.8
12.7
(spiro)
13.7
(spiro)
12.9
(spiro)
15.2
(spiro)
14.7
(spiro)
11.4
(spiro)
11.7
(spiro)
8.9
(spiro)
HTM-free
HTM-free
MS
MS
mpTiO2/MAPbI3/Au
mpTiO2/ZrO2/
(5-AVA)x(MA)1xPbI3/C
0.84
0.86
19
22.8
0.68
0.66
10.9
12.8
8.4 (spiro)
7.34
6.7
Remarks
Ref.
Spin-coating with
LiTFSI, TBP and FK209
Spin-coating with
LiTFSI and TBP
Spin-coating with
Doctor blading
Drop-casting
Low temperature
spattering
Electrodeposition
Solgel process
Spin-coating
35
Spin-coating with
LiTFSI and TBP
Spin-coating with
LiTFSI and D-TBP
Spin-coating with
LiTFSI and TBP
Electropolymerization
Spin-coating with
LiTFSI and TBP
Spin-coating
83
5
91
132
78
141
143
139
92
131
29
96
99
57
Spin-coating with
Spin-coating with
Spin-coating with
Spin-coating with
LiTFSI and TBP
Spin-coating with
Spin-coating
101
108
Spin-coating
109
Spin-coating with PDMS
133
102
104
105
106
114
120
MS refers to devices with mesoscopic architecture and P refers to devices with planar architecture.
8956 | J. Mater. Chem. A, 2015, 3, 89438969
View Article Online
Review
the greater possibility of fabricating exible solar cells on

polyethylene terephthlate (PET) substrates compared to their
mesoscopic counterparts. By taking advantage of low-temperature processibility, planar perovskite devices can also be integrated into tandem solar cells as top and/or bottom cell. In
addition, the scaold-less planar architecture oers considerable versatility in deposition techniques of organometal halide
perovskites, which further widens the options of materials as
charge-selective transporting (or blocking) layers to couple with
the respective perovskite. The other merit of PHJ solar cells is
that they can be employed as test beds to verify the ecacy of
novel materials owing to their simple and versatile preparation.
This section provides a summary of reports on various materials, both organic and inorganic, employed as charge-selective
transporting (or blocking) layers and interlayers in two types of
planar-heterojunction perovskite solar cells, i.e., conventional
and inverted architectures (Fig. 5).
3.2.1 Conventional nip planar perovskite solar cells.
The nip planar architecture strongly inherits the features of
the most commonly used mesoscopic perovskite devices (FTO/
TiO2/perovskite/spiro-OMeTAD/Au), hence the term conventional. Essentially an n-type electron selective layer is deposited on a transparent conductive substrate, followed by
sequential coating of hybrid perovskite layer, p-type hole
selective layer and top metal contact (light rst passes through
the bottom electron selective layer). The main dierence with
the corresponding mesoscopic devices is the use of a single
compact n-type metal oxide layer rather than a combination of
both compact and mesostructured scaolds. Planar perovskite
solar cells with such conguration have been shown to boast
high eciencies, comparable to their mesoscopic
counterparts.5,17,18
Electron-transporting materials. The mesoporous TiO2 (300
nm) layer is oen reported to be essential for high eciency
perovskite solar cells. This requirement may not seem as critical
aer looking at the result of substituting this mp-TiO2 layer with
a thinner layer with much lower porosity.8,35,99 Dar et al.
employed (001)-oriented anatase TiO2 nanoplatelets to replace
the mesoporous scaold and were able to obtain a PCE of 12.3%
aer thickness optimization of both the CH3NH3PbI3 and TiO2
electron-transporting layers (ETLs).125 This study implies that
under optimized fabrication conditions, reduced porosity has
marginal inuence on the achievable performance; and with
this in mind, planar devices with no porosity should also
perform reasonably.
TiO2 nanoparticles prepared using low-temperature solgel
synthesis can be used as an ETL for PHJ perovskite solar cells.126
A device with this ETL (ITO/TiO2/CH3NH3PbI3xClx/P3HT/Ag)
was able to achieve a PCE of 13.6% but this was strongly
dependent on the thermal history of the TiO2. The thermal
annealing of a compact TiO2 layer (<150 C) is proposed to
remove residual impurities and to improve the interparticle
connectivity. Nevertheless, prolonged heat treatment results in
a coarser morphology inducing highly detrimental eects on
device performance. Yella et al. also demonstrated a lowtemperature fabrication of TiO2/CH3NH3PbI3/spiro-OMeTAD
PHJ solar cells by utilizing nanocrystalline rutile TiO2 deposited
via chemical bath deposition from TiCl4 precursors.127 The

thickness and morphology of the rutile TiO2 nanoparticles can
be adjusted by controlling the concentration of the precursor
solution. Under optimized conditions, this rutile TiO2 PHJ
device resulted in an impressive PCE of 13.7% in contrast to
3.7% produced by anatase TiO2 prepared by spin-coating of
TiCl4 precursors. The rutile TiO2 device also exhibits an exceptionally high VOC of 1.11 V. This discrepancy is attributed to the
formation of a more intimate contact between rutile TiO2 and
the overlaying perovskite layer, resulting in a more eective
extraction of photogenerated electrons.
Besides TiO2, other n-type metal oxides (e.g., ZnO), which
have been applied in various mesoscopic perovskite devices,
can also be incorporated into planar solar cells.70,71 Liu and
Kelly reported the use of a low-temperature-processable ZnO
thin lm as an electron-transporting layer in planar CH3NH3PbI3 perovskite devices, demonstrating a PCE as high as 15.7%
under AM 1.5G illumination.18 The higher electron mobility of
ZnO compared to TiO2 renders it a better ETL material.128 The
ZnO nanoparticles prepared via hydrolysis of the Zn(O2CCH3)2
precursor require no thermal treatment and can be readily
deposited on conductive substrates. The planar devices fabricated are also highly reproducible with high VOC in excess of 1.0
V. The IIVI (e.g., CdSe) nanoparticles can also be used as an
electron-transporting layer in perovskite solar cells as demonstrated by Wang et al.129 The low-temperature solution-processed CdSe/CH3NH3PbI3/spiro-OMeTAD/Ag planar device
yielded a PCE of 11.7%.
Hole-transporting material. As discussed previously, organic
molecules, such as spiro-OMeTAD, are oen employed as a
hole-transporting material (HTM) for both dye-sensitized and
mesoscopic perovskite solar cells. Spiro-OMeTAD remains the
most popular choice as a HTM for conventional planar perovskite solar cells, which enables the attainment of devices with
high eciencies.5,17,18 Conings et al. demonstrated that a PCE as
high as 10.4% can be achieved from solar cells based on mixedhalide lead perovskites sandwiched between charge-selective
contacts of TiO2 and P3HT.130 The crystallinity of P3HT can be
improved with D-TBP instead of the commonly used TBP, as
reported by Guo et al. (a PCE of 9.2% for pristine P3HT and a
PCE of 12.4% for modied P3HT).131 Their study highlights that
semicrystalline polymers such as polythiophene derivatives are
energetically and electronically compatible with perovskite
materials, and can thus be alternatives to the amorphous small
molecules such as spiro-OMeTAD. Inorganic material CuSCN
was also applied as a HTM in a planar perovskite device
showing a PCE of 6.4%.132
Zheng et al. demonstrated another embodiment of planar
architecture by employing a highly hydrophobic oligothiophene
derivative (DR3TBDTT) as a HTM for PHJ perovskite solar cells
without any of the usual ionic additives such as Li-TFSI or
tBP.133 Instead, polydimethylsiloxane (PDMS) was used as a
morphological control agent to assist the formation of a
continuous and uniform HTM layer. The device showed decent
performance (PCE 8.8%) comparable to a device prepared
with spiro-OMeTAD + ionic additive (8.9%), with a much
J. Mater. Chem. A, 2015, 3, 89438969 | 8957
View Article Online
improved device stability owing to the moisture resistivity of the

oligothiophene molecule.
Interlayer. A few materials have been applied as interlayers in
conventional nip PHJ perovskite solar cells: ethoxylated
polyethylenimine (PEIE) and PC61BM. Such an interlayer is
expected to assist charge injection and extraction at metal
semiconductor interfaces, which in turn have signicant impact
on the device performance, via formation of a quasi-ohmic
interfacial contact. The strategy of incorporating an interlayer
has led to devices with high ll factors and low series resistance.
Zhou et al. engineered the interface by inserting a thin layer of
PEIE to reduce the work function of the ITO layer from 4.6 to 4.0
eV.5 It is possible to modify the work function with this interlayer because of the introduction of molecular dipole interactions and this leads to a better match with the conduction band
minimum (CBM) of the yttrium-doped TiO2 electron-transporting layer. In terms of device performance, the insertion of
the PEIE interlayer manifests itself by a higher photocurrent
and ll factor. Another example of interface engineering was
shown by Kim et al.134 A PC61BM layer was inserted between the
ZnO ETL and CH3NH3PbI3 layer and the presence of this
fullerene-derivative layer alters the electronic structure of the
ZnO layer, which eectively suppresses the recombination at
the ETLperovskite interface. This eventually leads to a PCE as
high as 12.2%, which is about 30% higher than devices without
any PC61BM interlayer.
3.2.2 Inverted pin planar perovskite solar cells. The
rst few papers employing inverted pin architecture are
based on a solution-processable organolead halide perovskite
sandwiched between a p-type conducting polymer (PEDOT:PSS)
and an n-type fullerene derivative (PC61BM), with a PEDOT:PSS
layer coated directly on the transparent conductive
substrate.15,33 The initial material selection was made with the
assumption that the perovskite has good p-type characteristics
and heavily inuenced by the pn heterojunction concept in
organic solar cells.135 PHJ devices with this architecture have
also experienced signicant performance enhancement in a
very short time, from a PCE of mere 3.9% to more than 15%.
Hole-transporting material. Some HTMs that have been used
in inverted planar-heterojunction perovskite solar cells include
PEDOT:PSS,15,33,136 polythiophene,137 spiro-OMeTAD,138 graphene oxide (GO),139 NiO,140143 V2O5,140 and CuSCN.141 Even
though PEDOT:PSS is one of the most commonly used HTMs for
inverted planar architecture, its work function is highly
dependent on the ratio of its ionomers (4.95.2 eV), and so
may not be high enough and may not perfectly match the
valence band maxima (VBM) of perovskite materials (e.g., 5.4
eV for CH3NH3PbI3). Lim et al. demonstrated that the
PEDOT:PSS work function could be tuned by adding a peruorinated ionomer (PFI) into the polymer solution.136 The PFI
was found to enrich the surface of PEDOT:PSS, resulting in a
deeper work function. In terms of device performance, the
interface engineering with PFI gave a PCE of 11.7%, more than
40% higher than the device without any PFI treatment.
An electrochemically polymerized polythiophene (PT) lm
was also used as a hole-transporting layer in ITO/PT/CH3NH3PbI3/C60/BCP/Ag PHJ solar cells due to its matching HOMO level
8958 | J. Mater. Chem. A, 2015, 3, 89438969
Review
(5.2 eV) with the VBM of perovskite materials.137 The devices

exhibit strong correlation between performance and PT lm
thickness, since the PT conductivity tends to increase with
thicker lms but its strong optical absorption may contribute to
loss of photons that can be absorbed by the perovskite layer.
Upon PT thickness optimization, a promising PCE of 11.8% was
obtained.
In an attempt to correlate the HOMO level of hole-transporting materials with the VOC of solar cells, Polander et al.
utilized inverted planar architecture with a series of HTMs
(HOMO levels ranging from 5.0 to 5.6 eV).138 Reducing
HOMO levels initially leads to an improved device performance,
particularly VOC, with the best device with spiro-TTB (HOMO
level 5.3 eV) having methyl groups instead of methoxy
groups on spiro-OMeTAD, demonstrating a PCE of 10.9% and
VOC of 1.03 V. HTMs with HOMO levels lower than 5.3 eV were
found to demonstrate poor performance due to the absence of
sucient driving force.
Graphene oxide (GO) is another alternative to PEDOT:PSS. As
demonstrated by Wu et al., GO lms were found to quench the
photoluminescence of CH3NH3PbI3xClx, suggesting the feasibility of charge transfer between the materials; the degree of
quenching varied depending on the GO thickness.139 Furthermore, due to the dierent surface chemistry, perovskite lms
formed on GO had better uniformity with fewer pores and
improved roughness as compared to those formed on
PEDOT:PSS. As a result, GO promotes a better surface coverage
and enhanced light scattering, which leads to a better PCE of
11.1% (vs. 9.3% for PEDOT:PSS). This is another indication that
promoting nucleation is important for good lm formation and
better coverage as GO provides a hydrophilic surface which
reduces the energy barrier for heterogeneous nucleation to
occur.
Since PEDOT:PSS is highly hygroscopic, which may
compromise the long-term stability of perovskite devices, it is
highly desirable to nd alternative materials; inorganic metal
oxides are ideal choices for this purpose since they are more
stable in nature. Docampo et al. were among the rst to apply
NiO as a hole-transporting layer in inverted planar devices but
without success, attributed mainly to poor perovskite lm
formation and poor surface coverage.140 Later development
showed that it is possible to achieve eciency as high as 9%
with NiO. There are various techniques employed to prepare a
NiO lm for PHJ perovskite solar cells: electrodeposition and
solgel synthesis.36,141143 Subbiah et al. used an electrodeposited
NiO lm with a vacuum-deposited hybrid perovskite lm (FTO/
NiO/CH3NH3PbI3xClx/PC61BM/Ag) and achieved a PCE of
7.3%.141 Jeng et al. applied a thin NiOx hole-transporting layer
and observed a better performance with a higher VOC compared
to devices prepared with PEDOT:PSS.142 UV-ozone treatment of
NiOx was found to improve surface wetting properties and
energy alignment between NiOx and perovskite, leading to more
than 1.5 times enhancement in device performance (PCE
7.8%). Similar conclusions on the eect of UV-ozone were
reached by Hu et al.36 By using nanocrystals prepared via a sol
gel method, Zhu et al. were able to obtain a compact, continuous and corrugated crystalline NiO lm with better charge
View Article Online
Review
extraction and transport properties compared to PEDOT:PSS.143

The surface texture of NiO assists the formation of large CH3NH3PbI3 crystals with improved light-scattering properties.
Consequently, following NiO thickness optimization, a PHJ
perovskite device with a PCE of 9.11% was achieved, much
higher as compared to devices with unstructured NiO thin lms
or with PEDOT:PSS.
Electron-transporting material. Fullerene (C60) and its derivatives are the most widely used n-type materials for ETLs in
inverted perovskite solar cells. It has been demonstrated that
PC61BM, one of the more commonly used fullerene derivatives,
could eectively quench the photoluminescence of CH3NH3PbI3, suggesting a high probability of ecient charge transfer.15
It remains debatable whether the presence of fullerene is
necessary to form the heterojunction required for dissociating
the photogenerated excitons formed in perovskite bulk lms or
whether the fullerene functions only as a charge extraction and
transport layer for the charge-carriers already formed in the
perovskite (i.e., fullerene acting as an electron acceptor with
reference to electron-donating perovskite). Dierent fullerene
derivatives typically possess distinctive solubilities in organic
solvents and dierent optoelectronic properties (e.g., electron
mobility and energy level). In terms of the device characteristics,
it is imperative to optimize the fullerene lm thickness to
maximize the attainable PCE. Seo et al. demonstrated that a
PC61BM layer as thick as 50 nm was sucient to attain full
coverage of the perovskite lm, while thicker fullerene layers
tended to increase the series resistance and modulate light
absorption.144 Apart from PC61BM, other fullerene compounds
have been used in similar device congurations: C60, ICBA and
PC71BM. Some of the ETL materials and a comparison of their
energy levels with the other layers in perovskite solar cells are
shown in Fig. 10 and 11. In a previous study by Jeng et al. on
inverted PHJ solar cells with fullerene derivatives, it was shown
that devices with PC61BM outperformed those prepared with
Fig. 11 Energy level diagram showing conduction band/LUMO levels

of various ETLs.
C60 and ICBA.33 Nevertheless, the correlation between the

LUMO level of the fullerenes and device photovoltage can be
established. The lower LUMO level of C60 compared to PC61BM
(4.5 eV vs. 3.9 eV) corresponds well to the VOC drop in the
solar cell devices. On the other hand, a cell prepared with ICBA,
which has a higher LUMO level than PC61BM (0.17 eV higher),
exhibits a higher VOC. Wang et al. also made similar observations in terms of the photovoltage trend with the three fullerene
derivatives in their planar devices; maximum VOC of 0.53, 0.91
and 1.06 V for C60, PC61BM and ICBA, respectively.52 By optimizing the deposition conditions of the perovskite lm and
incorporating additional C60 as an interlayer, they were able to
obtain a device with a PCE of 12.2% with ICBA. The use of
PC71BM to substitute PC61BM was demonstrated by Chiang
et al.22 The higher electron mobility and the better optical
absorption in the UV-visible range may have contributed to the
higher photocurrent in PC71BM devices. Solvent annealing of
the PC71BM layer, accomplished by keeping the sample in a
solvent-saturated environment, for 24 h was deemed necessary
to improve interfacial contact between PC71BM and perovskite
layers, as well as to enhance the PC71BM crystallinity. As a
Fig. 10 Structures of some of the ETL materials for perovskite solar cells.
J. Mater. Chem. A, 2015, 3, 89438969 | 8959
View Article Online
result, the optimized device employing the higher fullerene

exhibited 30% improvement in PCE, its best device performance with a PCE of 16.31%.
It is also desirable to nd alternative materials to the
fullerene family. Malinkiewicz et al. showed that organoborane
compounds (e.g., 3TPYMB) might be a suitable ETL for PHJ
perovskite solar cells.145 Devices fabricated with 3TPYMB
resulted in a PCE of 5.5% compared to 10% for PC61BM-based
devices. Nevertheless, it was also noted that the LUMO level of
3TPYMB was higher than that of PC61BM by 0.6 eV. Therefore,
there is an oset between the CBM of CH3NH3PbI3 and the
LUMO level of 3TPYMB that may inhibit electron transfer,
resulting in inferior JSC and VOC. It was also likely that the
suboptimal morphology of the organoborane layer contained
pinholes that might act as shunting pathways. Further netuning of the energy levels of organoborane as well as optimizing its lm morphology may result in more suitable ETL
candidates for organometal halide perovskites.
N-type metal oxides such as ZnO have also been tested in
planar devices with pin conguration. You et al. prepared
an ITO/PEDOT:PSS/CH3NH3PbI3xClx/ZnO/Al device by depositing ZnO nanoparticles from an orthogonal solvent (chlorobenzene). The device showed a remarkable eciency of 11.5%,
commensurate with that of PC61BM as an ETL, with external
quantum eciency peaking at 80%.146
Interlayer. There are two types of interlayers demonstrated
thus far for inverted planar architecture: p-type and n-type
interlayers (the former is to be inserted close to the holetransporting layer, while the latter is to be coupled with the
electron-transporting layer). Malinkiewicz et al. inserted a thin
layer (1020 nm) of organic polymer poly(N,N0 -bis(4-butylphenyl)-N,N0 -bis(phenyl)benzidine)
(polyTPD)
between
PEDOT:PSS and CH3NH3PbI3 layers using a meniscus-coating
process.145,147 The roles of polyTPD are two-fold: (1) to assist hole
transfer from CH3NH3PbI3 to PEDOT:PSS because the HOMO
level of polyTPD and the VBM of CH3NH3PbI3 are very similar in
energy (5.4 eV), and (2) to block the ow of electrons, which
also indirectly promotes the electron driing to the opposite
electrode, since the LUMO level of polyTPD (2.4 eV) is much
closer to vacuum compared with the CBM of CH3NH3PbI3 (3.9
eV). They further discovered that the conductivity of polyTPD
played a crucial role for thicker CH3NH3PbI3 lms.148 Perovskite
lms with various thicknesses were prepared (160 to 900 nm) by
a dual-source thermal evaporation method, providing high
control over lm uniformity and thickness accuracy, as well as
similar morphologies across the dierent lms. The best device
was obtained with a 285 nm thick perovskite lm (PCE
12.7%), while the 900 nm thick lm yielded poorer performance
(PCE 7.2%). Nevertheless, such thickness-dependent
behavior was not observed when the polyTPD interlayer was
slightly oxidized (or doped) with AgSbF6. With the polyTPD with
improved conductivity, the 900 nm thick CH3NH3PbI3 planar
device could generate PCE as high as 12%. The study also points
out that charge-carrier diusion lengths, measured to be in the
range of 100 nm in previous studies on CH3NH3PbI3, are
dependent on the perovskite lm preparation. It is also likely
that the high lm quality obtained via vacuum deposition
8960 | J. Mater. Chem. A, 2015, 3, 89438969
Review
signicantly improves the diusion lengths; hence, lm thickness no longer becomes the main limiting factor for the device
performance.
A number of materials have been demonstrated as functional n-type interlayers in pin planar heterojunction
devices, namely LiF,144 bathocuproine (BCP),33,52 bathophenanthroline (BPhen),38 ZnO,139 TiO2,140,149 Ca,38,150 C60 and its
derivatives32,52 and polyelectrolytes.151 Most of these compounds
have been previously applied as interlayers in organic solar
cells. They are typically incorporated between the electrontransporting layer (e.g., PC61BM) and the top electrode. The
insertion of thin LiF (0.5 nm), reported by Seo et al., results in
simultaneous improvement in both JSC and FF.144 The presence
of LiF generates dipole moment across the interfaces, which in
turn reduces the energy barrier between the LUMO level of
PC61BM and the Fermi level of Al, resulting in a better electron
extraction across both materials.
BCP is a widely used electron-blocking/hole-transporting
layer in organic light emitting diodes and organic solar cells.
There are a few roles assumed by BCP: (1) improving optical
eld; leading to increased absorption in the photoactive layer;
(2) blocking holes due to its poor hole mobility and its deep
HOMO level (7.0 eV); (3) transporting electrons due to its high
electron mobility and its matching energy (LUMO) level with
those of fullerene and electrode, and (4) protecting the fullerene
layer beneath from damage incurred from electrode deposition.
BPhen has similar characteristics to BCP, although the higher
electron mobility in BPhen may facilitate a more ecient
charge extraction and transport across the interlayer to the
electrode.152
Double fullerene layers, i.e., a stack of two layers of fullerene
or its derivatives, have been employed as an eective measure to
reduce current leakage. Wang et al. showed that the insertion of
a thermally evaporated C60 layer on a solution-deposited C60
dramatically reduced the dark current in the devices.52 This
indicates the crucial role of the additional C60 layer in preventing current leakage. In addition, a double fullerene C60/
PC61BM layer was found to passivate the defects (or reduce the
trap densities) at both the surface and grain boundaries of the
perovskite. A similar concept was also employed by Liang et al.
by using a solution-processable bis-C60 surfactant between the
PC61BM HTL and the top Ag electrode.32 The fullerene (bis-C60)
interlayer facilitates the energy alignment at the HTL/electrode
interface.
The possibility of exploiting the surface dipole has made
polyelectrolytes a suitable choice as an interlayer in planar
heterojunction solar cells. Zhang et al. applied two types of
solution-processable polyelectrolytes, i.e., PEIE and poly[3-(6trimethylammoniumhexyl)thiophene] (P3TMAHT), in devices
based on ITO/PEDOT:PSS/CH3NH3PbI3xClx/PC61BM/interlayer/Ag.151 Both polymers were found to decrease the work
function of Ag from 4.7 eV to 3.97 and 4.13 eV, for PEIE and
P3TMAHT, respectively, thereby lowering the electron injection
barrier at the PC61BM/Ag interface. Consequently, the performance of the devices improved from 8.5% (without interlayer)
to 12.0% (with PEIE) and 11.3% (with P3TMAHT).
View Article Online
Review
3.3
Mesoscopic versus planar architectures
Undeterred by their humble beginnings, solid-state organolead

halide hybrid perovskite solar cells with planar heterojunction
architecture have undergone a dramatic development over the
past few years. As discussed previously, up until now the optimization of thin-lm planar perovskite devices has led to
outstanding eciencies in excess of 15%, very comparable to or
in some cases even better than their mesoscopic equivalents,
thus opening the unavoidable discussion of the merits of the
planar and mesoscopic perovskite cells. The simplicity of the
planar heterojunction suggests that it is a more technologically
viable architecture to adopt and has very good large-scale
manufacturability. Leijtens et al. compared the electronic
properties of both architectures and found that the planar
perovskite lms were superior in terms of charge carrier
mobility (>20 cm2 V1 s1). However, solar cells prepared with
planar conguration suered from photovoltage loss, which
could be due to the high density of trap states in planar solar
cells resulting in high non-radiative electronhole recombination.153 Passivation or elimination of electronic trap sites in the
organicinorganic lead halide perovskites, for instance, via
Lewis base passivation as shown by Noel et al., can lead to
higher VOC in planar perovskite solar cells.21 Other than manufacturability and electrical mobility, there are other characteristics of the lms/devices that are inuenced by the
architecture of the cells. In the following sections, a comparison
between the impacts of the architectures on the dierent critical
aspects of the cells will be discussed.
3.3.1 Morphology and crystallization. As discussed in a
previous section, the perovskite lm morphology depends
strongly on the crystallization process involved. Compared to a
planar substrate, a mesoporous scaold with higher surface
area tends to promote a more extensive heterogeneous nucleation, which leads to the formation of smaller crystallites. Such
a distinction was observed by Ball et al.11 The CH3NH3PbI3xClx
perovskite lm prepared on a at TiO2 layer was found to have
crystals larger than 500 nm whereas the size of the crystals was
reduced to 60100 nm when deposited entirely within a mesoporous Al2O3 scaold (Fig. 12). Both polycrystalline lms exhibit
distinct absorption spectra; the spectra of the smaller
crystallites in the mesoporous scaold show broader and hypsochromically shied absorption spectra with respect to lms
deposited on the at substrates.154,155 Despite the larger crystal
size, which is generally more desirable for an ecient charge
transport, it is much more arduous to prepare excellent perovskite lms on planar substrates. Nevertheless, various methods
have been devised to meet this challenge, as summarized in the
previous morphology section. In the case of organolead halide
perovskites, highly ecient devices have been observed with
polycrystalline lms over a range of certain grain sizes, suggesting long charge-carrier diusion lengths and harmless grain
boundaries.17,35 In fact, as measured by Oga et al., the minimum
mobilities in CH3NH3PbI3 lms with various crystallite sizes
(70420 nm) range between 10 and 20 cm2 V1 s1.156 Apart from
the crystallite size, another concern that needs to be addressed
to enable production of devices with high eciencies includes
crystallite quality (presence of defects within or on the surface of
the crystals) and crystal orientation (specic direction for better
charge transport).
The other issue with mesoscopic cells that does not exist for
planar architecture is incomplete pore lling. The nanoscale
morphology of perovskite-lled TiO2 scaolds can be probed
visually by analytical TEM techniques, e.g., electron energy loss
spectroscopy (EELS) and electron spectroscopy imaging (ESI) as
was rst demonstrated by Nanova et al.157 Leijtens et al. pointed
out that incomplete pore lling of TiO2 scaolds with perovskite
materials could cause an unwanted contact between the p-type
HTM and n-type TiO2.68 By reducing the thickness of mesoporous TiO2 scaolds, they were able to improve the PCE from
6.3% to 8.6%, as improved pore lling could be attained and
following that the formation of a perovskite capping layer.
Other than optimizing the thickness and porosity of the scaffold, pore lling can be enhanced by improving the solution
viscosity, applying a longer inltration time prior to deposition
or by improving the surface properties of the scaold.
3.3.2 Film thickness, diusion length and charge-carrier
generation. One main disadvantage associated with planar
architecture proposed in the early stages of the development of
perovskite solar cells is that the device performance would be
critically constrained by exciton and free carrier diusion
Fig. 12 (a) Cross-sectional SEM micrographs of solar cells with (i) Al2O3 scaold and (ii) no scaold,11 (b) schematic representation of the average
crystal dimensions for samples deposited on an Al2O3 scaold (top) and on a at substrate (bottom). Reproduced with permission from ref. 154.
Copyright 2014 Nature Publishing Group.
J. Mater. Chem. A, 2015, 3, 89438969 | 8961
View Article Online
lengths. Early reports also suggested that perovskite solar cells

are excitonic in nature; this means that charge carriers are only
generated upon exciton dissociation across heterointerfaces.158
In the event of insuciently long exciton and free carrier
diusion lengths, a planar solar cell constructed with a thick
absorber lm, which is typically required to maximize photoabsorption, is expected to suer from poor performance. The
mesoscopic architecture oers a solution to this problem by
enabling maximum absorption without sacricing exciton and
charge recombination. Another solution is to use perovskite
materials with longer diusion lengths; CH3NH3PbI3xClx, for
instance, has electronhole diusion lengths in excess of 1 mm
while the triiodide (CH3NH3PbI3) perovskite has diusion
lengths of only 100 nm.13,14
Nevertheless, at present even CH3NH3PbI3 perovskite lms
with thickness approaching 1 mm have been employed in highly
ecient planar heterojunction solar cells.25,148 These results
raise the question of whether there is a need to reassess the
diusion lengths and further elucidate the working mechanism
in perovskite solar cells. As revealed by Edri et al., solar cells
based on perovskite sandwiched between hole- and electrontransporting layers operate as a pin device regardless of the
device architecture.87 There are two strong electron beaminduced current (EBIC) signals near the p and n regions,
suggesting strong charge separation and collection at both
interfaces with charge-selective transporting layers. The presence of the internal eld in the i region could assist chargecarrier dri. Docampo et al. proposed that the electronhole
diusion length could be dependent on the method employed
in the lm preparation, which consequently might aect the
lm morphology.23 They obtained diusion lengths of 200 nm
for CH3NH3PbI3 perovskite prepared by solution-based
sequential deposition. A much higher value of close to 1 mm was
estimated by Edri et al. in a separate EBIC study.123 They argued
that such variation could arise as their measurement was performed on the complete device, as opposed to only the thin lm,
in which the internal electric eld induced by selective contacts
might inuence the photoelectrical processes in the measured
sample. Given the low exciton binding energy of organolead
halide perovskites, the excitons may also dissociate into free
carriers in the bulk, liing the limitation imposed by exciton
diusion length.154 While there is still no consensus on the
exciton binding energy, it was found that free carriers accumulating at the band-edge of the hybrid perovskite CH3NH3PbI3
could promote a further reduction in the exciton binding energy
via coulombic screening of photogenerated excitons, which
enhances the formation of free carriers in the bulk.159 In fact, in
a theoretical investigation performed by Even et al., it was
suggested that the complete screening of the Wannier-Mott
excitons at room temperature, which was induced by the optical
phonons and collective rotational motion of the organic
cations, could lead to almost free carriers.160 To conclude this
section, the mesoscopic device is thus the preferred choice for
materials with short diusion lengths due to the smaller
distance the excitons/free carriers have to travel, which in turn
also depends on the pore sizes in the scaold.
8962 | J. Mater. Chem. A, 2015, 3, 89438969
Review
3.3.3 Hysteresis. Anomalous IV hysteresis phenomena,

i.e., variations in the IV characteristics depending on the
direction and the rate of voltage sweep, were observed for
perovskite solar cells and this posed a great concern as they
undermine the accuracy of the measured eciencies of these
devices.161 The photocurrent appears to be dependent on the
bias voltage applied just before the measurement is performed;
hence the device can be pre-conditioned. Therefore, a voltage
sweep in the direction of open circuit to short circuit tends to
articially enhance the current measured. The origin of
hysteresis is still open to debate, although defects and polarizability of the perovskite have been proposed as possible
reasons. The use of a mesoporous TiO2 scaold signicantly
impairs the hysteresis eects, while the corresponding planar
devices suer from severe hysteresis (Fig. 13).29,162164 Kim and
Park discovered that the low-frequency capacitance was higher
in a planar device, suggesting that a planar substrate could
induce dipole polarization in CH3NH3PbI3 perovskite lms,
which was considered to be the cause of the hysteresis.162 On the
other hand, Snaith et al. proposed that the severity of hysteresis
was dependent on the quality of interface between perovskite
and the charge-selective contacts.163 Due to the higher surface
area, charge transfer between perovskite and a mesoporous
TiO2 scaold is more ecient, resulting in diminished impact
of the phenomenon. Hysteresis could still be observed in cells
fabricated with a mesoporous passive insulating Al2O3 scaffold. Noel et al. demonstrated that the hysteresis of a planar
device could be reduced by passivating the surface defects of the
perovskite layer with treatments of pyridine or thiophene.21
We surmise that the device architecture is not the inherent
root cause of the hysteresis observed. While a mesoporous TiO2
scaold appears to be successful in attenuating hysteresis, it is
clear that this is not so for an Al2O3 scaold. Even the degree of
hysteresis varies depending on the thickness of the mesoporous
TiO2.163 Interestingly, inverted planar perovskite devices based
on PEDOT:PSS/perovskite/PC61BM with high eciencies have
been reported without any observed hysteresis.19,22,25 We also
note that the photoluminescence quenching of CH3NH3PbI3xClx with PEDOT:PSS and PC61BM is more ecient than
with spiro-OMeTAD and TiO2, respectively.140 This can be
translated into a more ecient hole transfer from the perovskite layer to PEDOT:PSS than to spiro-OMeTAD and a more
ecient electron transfer from the perovskite layer to PC61BM
than to TiO2. Extrapolating from the conclusion given by Snaith
et al. on the impact of interface quality, we believe that the
inverted perovskite device with this material combination is
less likely to develop hysteresis and that a hysteresis-free could
be even obtained from a conventional nip device with the
appropriate materials as charge-selective contacts.163 As such,
the hysteresis issue could be overcome by proper material
selection and, to a certain extent, morphology optimization
notwithstanding the type of device architecture.
3.3.4 PbI2 contribution. It has been previously demonstrated that PbI2 formed upon the decomposition of organolead
halide perovskites.53 For reasons that are not yet fully understood, the formation of a PbI2 phase is more prominent in
View Article Online
Review
Fig. 13 Cross-sectional SEM images of CH3NH3PbI3 devices with (a) 220 and (b) 110 nm thick mp-TiO2 and (c) planar device without mp-TiO2.
IV curves measured at FS (solid line) and RS (dashed line) for mesostructured devices with (d) 220 and (e) 110 thick mp-TiO2 and (f) planar device
without mp-TiO2. Reproduced with permission from ref. 162. Copyright 2014 American Chemical Society.
hybrid perovskite samples prepared on a mesoporous scaold

than on a planar substrate.155 Due to its wide bandgap (Eg 2.3
eV), the presence of PbI2 species is commonly undesirable
because of its poorer optical absorption properties than those of
CH3NH3PbI3. Mesoscopic solar cell devices prepared with PbI2
(FTO/c-TiO2/mp-TiO2/PbI2/spiro-OMeTAD/Ag) have demonstrated poor performance, attributing to its signicantly inferior transport properties (smaller diusion coecient) to those
of CH3NH3PbI3-based devices.165 PbI2 forming in a perovskite
lm, as in the case of an over-annealed perovskite lm, might
cause energy misalignment at the TiO2 interface and it could
also trap free charges generated in the perovskite phase, leading
to a deteriorated device.
The notion of a detrimental PbI2 phase is not shared by

everyone, as it has also been proposed that a tiny amount of PbI2
can lead to self-passivation of the defect states in the perovskite
lm and charge selectivity at the perovskite interface.166,167 Chen
et al. observed the formation of PbI2 species along the grain
boundaries, although extended annealing resulted in the
conversion of whole perovskite grains.108 There are three plausible roles of PbI2 in self-passivating CH3NH3PbI3: (1) to
preclude excitons from surface defects due to the type I energy
alignment; (2) to reduce recombination between electrons from
TiO2 and holes from perovskite at interface I, and (3) to reduce
recombination between electrons from perovskite and holes
from the hole-transporting material (HTM) (Fig. 14). The presence of a small amount of PbI2 in perovskite lms also increases
Fig. 14 Schematics of the proposed mechanism of self-passivation of PbI2 in a CH3NH3PbI3 lm. Reproduced with permission from ref. 108.
Copyright 2014 American Chemical Society.
J. Mater. Chem. A, 2015, 3, 89438969 | 8963
View Article Online
carrier lifetimes, leading to improved device performances.108,155

As compared to planar devices, mesoscopic perovskite solar
cells may be more easily self-passivated due to the higher
chance for PbI2 formation. However, this may also lead to
adverse eects if too much PbI2 phase is present, imposing
more stringent requirements on the processing of these thin
lms.
4. Material and device stability

Perovskite solar cells have undoubtedly demonstrated the
possibility of attaining excitingly high eciencies comparable
to those obtained for thin-lm solar cells in the foreseeable
future. It is still confronted by a very crucial challenge before
any worldwide acceptance and commercial viability can be
established, and that is the long-term environmental stability
(air, moisture and thermal stability) of the devices that is largely
dependent on the choice of materials. Organometal halide
perovskites are intrinsically susceptible to moisture and heat,
while the organics in the cells tend to degrade in the presence of
moisture, oxygen and heat. The instability of lead halide
perovskites (e.g., CH3NH3PbI3) in a humid environment is oen
reected by the decrease in lm absorption between 530 and
800 nm, as reported by Niu et al.168 The moisture-induced
degradation of CH3NH3PbI3 is also evident from the disappearance of the XRD peaks related to the initial perovskite
structure and the emergence of peaks corresponding to the
degradation products comprising PbI2 and I2; the formation of
I2 is triggered only when both moisture and light are present.
Habisreutinger et al. also pointed out that the color change
from dark brown/black to yellow in lead halide perovskite lms
upon degradation in air could be attributed to the bond dissociation among the crystal units induced by the interaction
between water molecules and the highly hygroscopic methylammonium cations, resulting in the transformation of the
perovskite crystal structure to a lower dimensional system.95
Therefore, perovskite solar cells require stringent encapsulation
to ensure a long outdoor operational lifetime.
Despite the pressing issue and the great interest in device
stability, the number of published studies on long-term device
stability remains scarce. The rst work on long-term stability of
perovskite solar cells was reported by Kim et al. demonstrating
stable performance over 500 h, although the study was only
done ex situ, i.e., devices were stored in air at room temperature
and were only intermittently measured under AM 1.5G illumination.8 The precise conditions for ageing studies of perovskite
solar cells remain unclear and therefore need to be standardized. To resolve this ambiguity, Gr
atzel recommended longterm stability tests for encapsulated devices that include lightsoaking tests (1000 h in full sunlight) and damp heat tests (1000
h at 85%RH and 85 C).169 A highly promising long-term device
stability was obtained by Burschka et al. demonstrating an
encapsulated mesoscopic perovskite solar cell (FTO/c-TiO2/mpTiO2/CH3NH3PbI3/spiro-OMeTAD/Au) that only degraded to
approximately 80% of its initial eciency aer being lightsoaked (100 mW cm2 at 45 C) for 500 h.35 The absence of any
signicant change in the device photocurrent aer the testing
8964 | J. Mater. Chem. A, 2015, 3, 89438969
Review
period suggests that the perovskite material remains photochemically stable under encapsulation. Law et al. also demonstrated that sealed perovskite solar cells with similar
architecture only suered from an 8% decrease in photocurrent
under continuous high-intensity illumination (equivalent to
40 sun) for >60 h.170 Nevertheless, a much poorer stability was
observed upon substitution of spiro-OMeTAD with the other
HTMs (e.g., P3HT and DPPTTT), replacement of the mp-TiO2
scaold with mp-Al2O3 and when encapsulation was not
employed.
Leitjens et al. argued that the stability study had to be realized under simulated solar illumination instead of white light
emitting diodes (LEDs) that was devoid of ultraviolet (UV)
components,171 under which the abovementioned studies by
Burschka et al. and Law et al. were performed.35,170 They
discovered that under AM1.5G (1 sun) illumination the encapsulated FTO/c-TiO2/mp-TiO2/CH3NH3PbI3xClx/spiro-OMeTAD/
Au device degraded to <10% of its initial eciency within 5 h,
while the corresponding UV-ltered device decayed to only
85% of its initial performance. Oxygen molecules in air tend
to be adsorbed on the oxygen vacancies on the TiO2 surface. It
was hypothesized that upon UV light excitation, the photogenerated holes in the valence band could recombine with the
electrons at oxygen vacancies at the TiO2 surface, leaving free
electrons in the conduction band and generating unlled
oxygen vacancy sites (Fig. 15). These deep electronic sites could
further trap the photogenerated electrons, from which they
could recombine with holes in the HTM, which results in the
instability of the UV-aged perovskite solar cells. To circumvent
the inherent UV-induced instability of TiO2-based solar cells,
Leitjens et al. proposed the use of insulating Al2O3 to substitute
the TiO2 scaold. Expectedly, the TiO2-free system demonstrated a much enhanced stability with a decrease of 50% of
its initial eciency within the rst 200 h of exposure and an
impressive stabilized PCE at 6% over 1000 h of continuous AM
1.5G illumination without any UV lter.
Schematics of the proposed mechanism of UV-induced

degradation in a TiO2-based solar cell. Reproduced with permission
from ref. 171. Copyright 2013 Nature Publishing Group.
Fig. 15
View Article Online
Review
The long-term stability of perovskite solar cells can also be

improved by employing additional barrier layers to protect the
perovskite lm beneath, in addition to external encapsulation
strategies. The ecacy of such protection largely depends on
gas permeability and hydrophobicity of the moisture-shielding layer. Habisreutinger et al. substituted organic HTMs with
a carbon nanotubePMMA composite, demonstrating an
enhanced stability of the protected organolead halide perovskite of up to 100 h at 80 C.95 This is evident from the lack of
change in the absorption spectra and crystal structure of the
perovskite since the hydrophobic PMMA is expected to prevent
the permeation of water molecules. Improvement in moisture
stability was also attained upon insertion of a thin Al2O3 barrier
layer between CH3NH3PbI3 and spiro-OMeTAD layers.168 Upon
exposure to humidity of 60% at 35 C, the device with Al2O3
incorporation demonstrated a drop of 50% of its initial eciency in contrast to a decrease of 80% within 18 h. Moisture
barrier can also be extended by employing a moisture-impermeable counter electrode even in the absence of the HTM layer.
The use of a thick carbon electrode (10 mm) in the HTM-free
perovskite solar cell, discussed previously in Section 3.1.4,
remarkably improved its long-term stability, i.e., no apparent
decrease in device PCE, under AM 1.5G illumination in air over
1000 h without any encapsulation.120 The other strategy to
improve moisture stability is by using HTMs with higher
hydrophobicity, thus preventing water inltration into the
perovskite layer. In an ex situ stability test at 20% humidity, an
unsealed perovskite solar cell with a HTM based on a hydrophobic polymer was shown to maintain its PCE stable for 1000
h, while the corresponding device with spiro-OMeTAD as its
HTM showed 28% reduction in PCE.100 Device stability can be
further improved by eliminating the incorporation of highly
hygroscopic ionic additives, e.g., Li-TFSI, commonly expected to
improve the hole mobility, in conjunction with the use of a
hydrophobic HTM with high mobility.133 Furthermore, the
choice of a HTM additive is also expected to aect the long-term
stability of perovskite solar cells.172
Besides moisture stability, thermal stability is another
fundamental concern since continuous exposure to light illumination (e.g., 100 mW cm2 in full sun) may heat up the solar
cells, raising their operating temperature to as high as 80 C.173
This may exacerbate any moisture-induced degradation
processes already present in the devices. Organolead halide
hybrid perovskites have been reported to degrade into their
original lead halide precursors as the more unstable organohalide components decompose under heating. Although metal
halide perovskite materials have been reported to be stable
above 300 C,92,174,175 recent work demonstrated that the
decomposition of the organic component occurs at temperatures much lower than 300 C, even as low as 140 C.53,166,176 As
discussed previously in Section 3.3.4, the presence of metal
halide (e.g., PbI2) as a degradation product of the hybrid
perovskite (e.g., CH3NH3PbI3) may interfere with the electrical
processes in perovskite solar cells due to their inferior electrical
properties and incompatible bandgap alignment. These lead
halides may also possess poor photon harvesting capability. On
the other hand, the thermal decomposition of organohalide
(e.g., CH3NH3I) may lead to the formation of a hydrogen halide

(e.g., HI) and an amine (e.g., CH3NH2); the latter tends to remain
more persistently within the perovskite matrix, subsequently
aecting the photoconversion processes.176 The thermal
stability of perovskite solar cells can be improved by either
employing impervious barrier lms to prevent the organic
components from escaping or utilizing new perovskite materials with better inherent thermal stability.95,177
In summary, other than competitive eciencies, robust solar
cells with excellent moisture and thermal stabilities are highly
sought aer. As discussed in the previous section, the presence
of mild moisture can in fact assist the formation of a perovskite lm with larger grain size and reduced pinholes.5,57
Dierent approaches to improve the device stability developed
thus far have mainly emphasized on the incorporation of
barrier lms in various forms. Nevertheless, it is more crucial,
in our opinion, to develop materials with novel chemistries that
are less susceptible to moisture and prolonged heating. The use
of a mixed halide, for instance, is a viable method to engineer
perovskite materials with enhanced stability.178 Device architecture, which remains ambiguous in terms of its eects on the
device stability, should also be considered when performing
future long-term stability tests.
5.
Summary
The comparison of the merits of each of the architectures is

inevitable as mesoscopic and planar perovskite devices, derived
from very dierent origins (dye-sensitized solar cells vs. thinlm solar cells), have achieved similarly excellent eciencies.
The reason for such universal good performances is most
probably the forgiving optoelectronic properties oered by this
class of materials (high absorption coecient, bandgap
compatible with the solar spectrum, excellent charge lifetimes
and good charge mobilities). That being said, the rapid development of this eld only took o in 2013 and the reason for this
is that there was then a common recognition and acceptance of
the importance of controlled crystallization in these lms.
Hence in this review, the two important factors contributing to
the spectacular or, in some cases, the lack of spectacular
performance in perovskite solar cells morphology of the
perovskite lms and the architecture of the cells are discussed.
Many dierent physical and chemical methods have been used
to control the crystallization process with various degrees of
success. The better control has led to reasonable performance
achieved in both architectures. That being said, mesoporous
architectures provide a simple and straightforward way to achieve controllable crystallization with good surface nish.
However, the preparation of the mesoporous scaold also
requires a careful selection of materials, hence limiting the
material options available for this type of architecture. Due to
their low-temperature processability, planar heterojunction
perovskite solar cells are better suited for fabrication on exible
polymer substrates. Thus far exible planar devices have
demonstrated high eciencies approaching 10%.18,136,146
Commercially attractive semitransparent solar cells can be
more easily implemented with planar conguration as solar
J. Mater. Chem. A, 2015, 3, 89438969 | 8965
View Article Online
windows for buildings and vehicles. Perovskite materials are

ideal for this purpose due to their high optical absorption even
with reasonably thin lms and tunability of the optical transmittance by controlling the lm morphology.179,180 The other key
advantage of planar heterojunctions over mesoscopic architectures is their simple fabrication, rendering it a more viable
option as the platform to evaluate the potential of novel materials and more intricate multi-junction architectures.181,182 The
high charge mobility and versatility oered by planar architectures is oset by the limit in the photovoltage due to the presence of sub-gap states and low intrinsic doping densities.153 For
highly ecient solar cells, a combination of both planar and
mesoscopic architecture may eventually be the way to go
forward. Regardless of the architecture forward, lowering the
energy for nucleation and controlled growth in these lms
should lead to better solar cell performance.
References
1 K. Chondroudis and D. B. Mitzi, Chem. Mater., 1999, 11,
30283030.
2 D. B. Mitzi, Prog. Inorg. Chem., 1999, 48, 1121.
3 D. B. Mitzi, C. A. Feild, Z. Schlesinger and R. B. Laibowitz, J.
Solid State Chem., 1995, 114, 159163.
4 A. Kojima, K. Teshima, Y. Shirai and T. Miyasaka, J. Am.
Chem. Soc., 2009, 131, 60506051.
5 H. Zhou, Q. Chen, G. Li, S. Luo, T.-b. Song, H.-S. Duan,
Z. Hong, J. You, Y. Liu and Y. Yang, Science, 2014, 345,
542546.
6 J.-H. Im, C.-R. Lee, J.-W. Lee, S.-W. Park and N.-G. Park,
Nanoscale, 2011, 3, 40884093.
7 M. M. Lee, J. Teuscher, T. Miyasaka, T. N. Murakami and
H. J. Snaith, Science, 2012, 338, 643647.
8 H.-S. Kim, C.-R. Lee, J.-H. Im, K.-B. Lee, T. Moehl,
A. Marchioro, S.-J. Moon, R. Humphry-Baker, J.-H. Yum,
J. E. Moser, M. Gr
atzel and N.-G. Park, Sci. Rep., 2012, 2,
DOI: 10.1038/srep00591.
9 M. A. Green, J. Mater. Sci.: Mater. Electron., 2007, 18, 1519.
10 P. Kovacik, G. Sforazzini, A. G. Cook, S. M. Willis,
P. S. Grant, H. E. Assender and A. A. R. Watt, ACS Appl.
Mater. Interfaces, 2010, 3, 1115.
11 J. M. Ball, M. M. Lee, A. Hey and H. J. Snaith, Energy Environ.
Sci., 2013, 6, 17391743.
12 J. M. Frost, K. T. Butler, F. Brivio, C. H. Hendon, M. van
Schilfgaarde and A. Walsh, Nano Lett., 2014, 14, 25842590.
13 S. D. Stranks, G. E. Eperon, G. Grancini, C. Menelaou,
M. J. P. Alcocer, T. Leijtens, L. M. Herz, A. Petrozza and
H. J. Snaith, Science, 2013, 342, 341344.
14 G. Xing, N. Mathews, S. Sun, S. S. Lim, Y. M. Lam,
M. Gr
atzel, S. Mhaisalkar and T. C. Sum, Science, 2013,
342, 344347.
15 S. Sun, T. Salim, N. Mathews, M. Duchamp, C. Boothroyd,
G. Xing, T. C. Sum and Y. M. Lam, Energy Environ. Sci.,
2014, 7, 399407.
16 W.-J. Yin, T. Shi and Y. Yan, Adv. Mater., 2014, 26, 4653
4658.
8966 | J. Mater. Chem. A, 2015, 3, 89438969
Review
17 M. Liu, M. B. Johnston and H. J. Snaith, Nature, 2013, 501,

395398.
18 D. Liu and T. L. Kelly, Nat. Photonics, 2014, 8, 133138.
19 Z. Xiao, C. Bi, Y. Shao, Q. Dong, Q. Wang, Y. Yuan, C. Wang,
Y. Gao and J. Huang, Energy Environ. Sci., 2014, 7, 2619
2623.
20 M. Xiao, F. Huang, W. Huang, Y. Dkhissi, Y. Zhu,
J. Etheridge, A. Gray-Weale, U. Bach, Y.-B. Cheng and
L. Spiccia, Angew. Chem., Int. Ed., 2014, 53, 98989903.
21 N. K. Noel, A. Abate, S. D. Stranks, E. Parrott, V. Burlakov,
A. Goriely and H. J. Snaith, ACS Nano, 2014, 8, 98159821.
22 C.-H. Chiang, Z.-L. Tseng and C.-G. Wu, J. Mater. Chem. A,
2014, 2, 1589715903.
23 P. Docampo, F. Hanusch, S. D. Stranks, M. D
oblinger,
J. M. Feckl, M. Ehrensperger, N. K. Minar, M. B. Johnston,
H. J. Snaith and T. Bein, Adv. Energy Mater., 2014, 4,
1400355.
24 F. Huang, Y. Dkhissi, W. Huang, M. Xiao, I. Benesperi,
S. Rubanov, Y. Zhu, X. Lin, L. Jiang, Y. Zhou, A. GrayWeale, J. Etheridge, C. R. McNeill, R. A. Caruso, U. Bach,
L. Spiccia and Y.-B. Cheng, Nano Energy, 2014, 10, 1018.
25 Z. Xiao, Q. Dong, C. Bi, Y. Shao, Y. Yuan and J. Huang, Adv.
Mater., 2014, 26, 65036509.
26 S. Gamliel and L. Etgar, RSC Adv., 2014, 4, 2901229021.
27 M. He, D. Zheng, M. Wang, C. Lin and Z. Lin, J. Mater.
Chem. A, 2014, 2, 59946003.
28 G. E. Eperon, V. M. Burlakov, P. Docampo, A. Goriely and
H. J. Snaith, Adv. Funct. Mater., 2014, 24, 151157.
29 N. J. Jeon, J. H. Noh, Y. C. Kim, W. S. Yang, S. Ryu and
S. I. Seok, Nat. Mater., 2014, 13, 897903.
30 J. C. Maxwell, Philos. Mag., 1908, 16, 818824.
31 D. Yang, M. Krasowska, R. Sedev and J. Ralston, Phys. Chem.
Chem. Phys., 2010, 12, 1372413729.
32 P.-W. Liang, C.-Y. Liao, C.-C. Chueh, F. Zuo, S. T. Williams,
X.-K. Xin, J. Lin and A. K. Y. Jen, Adv. Mater., 2014, 26, 3748
3754.
33 J.-Y. Jeng, Y.-F. Chiang, M.-H. Lee, S.-R. Peng, T.-F. Guo,
P. Chen and T.-C. Wen, Adv. Mater., 2013, 25, 37273732.
34 Y.-F. Chiang, J.-Y. Jeng, M.-H. Lee, S.-R. Peng, P. Chen,
T.-F. Guo, T.-C. Wen, Y.-J. Hsu and C.-M. Hsu, Phys.
Chem. Chem. Phys., 2014, 16, 60336040.
35 J. Burschka, N. Pellet, S.-J. Moon, R. Humphry-Baker,
P. Gao, M. K. Nazeeruddin and M. Gr
atzel, Nature, 2013,
499, 316319.
36 L. Hu, J. Peng, W. Wang, Z. Xia, J. Yuan, J. Lu, X. Huang,
W. Ma, H. Song, W. Chen, Y.-B. Cheng and J. Tang, ACS
Photonics, 2014, 1, 547553.
37 H. Hu, D. Wang, Y. Zhou, J. Zhang, S. Lv, S. Pang, X. Chen,
Z. Liu, N. P. Padture and G. Cui, RSC Adv., 2014, 4, 28964
28967.
38 C.-W. Chen, H.-W. Kang, S.-Y. Hsiao, P.-F. Yang,
K.-M. Chiang and H.-W. Lin, Adv. Mater., 2014, 26, 6647
6652.
39 Q. Chen, H. Zhou, Z. Hong, S. Luo, H.-S. Duan, H.-H. Wang,
Y. Liu, G. Li and Y. Yang, J. Am. Chem. Soc., 2013, 136, 622
625.
View Article Online
Review
40 H.-B. Kim, H. Choi, J. Jeong, S. Kim, B. Walker, S. Song and

J. Y. Kim, Nanoscale, 2014, 6, 66796683.
41 Y. Wu, A. Islam, X. Yang, C. Qin, J. Liu, K. Zhang, W. Peng
and L. Han, Energy Environ. Sci., 2014, 7, 29342938.
42 Y. Zhao and K. Zhu, J. Phys. Chem. C, 2014, 118, 94129418.
43 H. Yu, F. Wang, F. Xie, W. Li, J. Chen and N. Zhao, Adv.
Funct. Mater., 2014, 24, 71027108.
44 P.-W. Liang, C.-C. Chueh, X.-K. Xin, F. Zuo, S. T. Williams,
C.-Y. Liao and A. K. Y. Jen, Adv. Energy Mater., 2014, 4,
1400960.
45 Y. Zhao and K. Zhu, J. Am. Chem. Soc., 2014, 136, 12241
12244.
46 C. Zuo and L. Ding, Nanoscale, 2014, 6, 99359938.
47 K. W. Tan, D. T. Moore, M. Saliba, H. Sai, L. A. Estro,
T. Hanrath, H. J. Snaith and U. Wiesner, ACS Nano, 2014,
8, 47304739.
48 Y. Tidhar, E. Edri, H. Weissman, D. Zohar, G. Hodes,
D. Cahen, B. Rybtchinski and S. Kirmayer, J. Am. Chem.
Soc., 2014, 136, 1324913256.
49 S. T. Williams, F. Zuo, C.-C. Chueh, C.-Y. Liao, P.-W. Liang
and A. K. Y. Jen, ACS Nano, 2014, 8, 1064010654.
50 M. H. Du, J. Mater. Chem. A, 2014, 2, 90919098.
51 E. Mosconi, E. Ronca and F. De Angelis, J. Phys. Chem. Lett.,
2014, 5, 26192625.
52 Q. Wang, Y. Shao, Q. Dong, Z. Xiao, Y. Yuan and J. Huang,
Energy Environ. Sci., 2014, 7, 23592365.
53 A. Dualeh, N. T
etreault, T. Moehl, P. Gao,
M. K. Nazeeruddin and M. Gr
atzel, Adv. Funct. Mater.,
2014, 24, 32503258.
54 M. Saliba, K. W. Tan, H. Sai, D. T. Moore, T. Scott,
W. Zhang, L. A. Estro, U. Wiesner and H. J. Snaith, J.
Phys. Chem. C, 2014, 118, 1717117177.
55 H.-L. Hsu, C.-P. Chen, J.-Y. Chang, Y.-Y. Yu and Y.-K. Shen,
Nanoscale, 2014, 6, 1028110288.
56 G. Li, Y. Yao, H. Yang, V. Shrotriya, G. Yang and Y. Yang,
Adv. Funct. Mater., 2007, 17, 16361644.
57 J. You, Y. Yang, Z. Hong, T.-B. Song, L. Meng, Y. Liu,
C. Jiang, H. Zhou, W.-H. Chang, G. Li and Y. Yang, Appl.
Phys. Lett., 2014, 105, 183902.
58 J. H. Rhee, C.-C. Chung and E. W.-G. Diau, NPG Asia Mater.,
2013, 5, e68.
59 N.-G. Park, J. Phys. Chem. Lett., 2013, 4, 24232429.
60 H.-S. Kim, S. H. Im and N.-G. Park, J. Phys. Chem. C, 2014,
118, 56155625.
61 B. Wang, X. Xiao and T. Chen, Nanoscale, 2014, 6, 12287
12297.
62 H. J. Snaith, J. Phys. Chem. Lett., 2013, 4, 36233630.
63 P. P. Boix, K. Nonomura, N. Mathews and S. G. Mhaisalkar,
Mater. Today, 2014, 17, 1623.
64 U. Bach, D. Lupo, P. Comte, J. E. Moser, F. Weissortel,
J. Salbeck, H. Spreitzer and M. Gr
atzel, Nature, 1998, 395,
583585.
65 T. C. Sum and N. Mathews, Energy Environ. Sci., 2014, 7,
25182534.
66 P. Gao, M. Gr
atzel and M. K. Nazeeruddin, Energy Environ.
Sci., 2014, 7, 24482463.
67 A. Mei, X. Li, L. Liu, Z. Ku, T. Liu, Y. Rong, M. Xu, M. Hu,

J. Chen, Y. Yang, M. Gr
atzel and H. Han, Science, 2014,
345, 295298.
68 T. Leijtens, B. Lauber, G. E. Eperon, S. D. Stranks and
H. J. Snaith, J. Phys. Chem. Lett., 2014, 5, 10961102.
69 D. Bi, G. Boschloo, S. Schwarzmuller, L. Yang,
E. M. J. Johansson and A. Hagfeldt, Nanoscale, 2013, 5,
1168611691.
70 M. H. Kumar, N. Yantara, S. Dharani, M. Graetzel,
S. Mhaisalkar, P. P. Boix and N. Mathews, Chem.
Commun., 2013, 49, 1108911091.
71 D.-Y. Son, J.-H. Im, H.-S. Kim and N.-G. Park, J. Phys. Chem.
C, 2014, 118, 1656716573.
72 X. Gao, J. Li, J. Baker, Y. Hou, D. Guan, J. Chen and C. Yuan,
Chem. Commun., 2014, 50, 63686371.
73 J. Qiu, Y. Qiu, K. Yan, M. Zhong, C. Mu, H. Yan and S. Yang,
Nanoscale, 2013, 5, 32453248.
74 K. Manseki, T. Ikeya, A. Tamura, T. Ban, T. Sugiura and
T. Yoshida, RSC Adv., 2014, 4, 96529655.
75 D. Zhong, B. Cai, X. Wang, Z. Yang, Y. Xing, S. Miao,
W.-H. Zhang and C. Li, Nano Energy, 2015, 11, 409418.
76 H. Tian, B. Xu, H. Chen, E. M. J. Johansson and
G. Boschloo, ChemSusChem, 2014, 7, 21502153.
77 K.-C. Wang, J.-Y. Jeng, P.-S. Shen, Y.-C. Chang,
E. W.-G. Diau, C.-H. Tsai, T.-Y. Chao, H.-C. Hsu, P.-Y. Lin,
P. Chen, T.-F. Guo and T.-C. Wen, Sci. Rep., 2014, 4, DOI:
10.1038/srep04756.
78 K.-C. Wang, P.-S. Shen, M.-H. Li, S. Chen, M.-W. Lin,
P. Chen and T.-F. Guo, ACS Appl. Mater. Interfaces, 2014,
6, 1185111858.
79 L. Etgar, P. Gao, P. Qin, M. Graetzel and M. K. Nazeeruddin,
J. Mater. Chem. A, 2014, 2, 1158611590.
80 A. Abrusci, S. D. Stranks, P. Docampo, H.-L. Yip, A. K. Y. Jen
and H. J. Snaith, Nano Lett., 2013, 13, 31243128.
81 Z. Zhu, J. Ma, Z. Wang, C. Mu, Z. Fan, L. Du, Y. Bai, L. Fan,
H. Yan, D. L. Phillips and S. Yang, J. Am. Chem. Soc., 2014,
136, 37603763.
82 J. T.-W. Wang, J. M. Ball, E. M. Barea, A. Abate,
J. A. Alexander-Webber, J. Huang, M. Saliba, I. Mora-Sero,
J. Bisquert, H. J. Snaith and R. J. Nicholas, Nano Lett.,
2013, 14, 724730.
83 K. Wojciechowski, M. Saliba, T. Leijtens, A. Abate and
H. J. Snaith, Energy Environ. Sci., 2014, 7, 11421147.
84 D. Bi, S.-J. Moon, L. Haggman, G. Boschloo, L. Yang,
E. M. J. Johansson, M. K. Nazeeruddin, M. Gr
atzel and
A. Hagfeldt, RSC Adv., 2013, 3, 1876218766.
85 S. H. Hwang, J. Roh, J. Lee, J. Ryu, J. Yun and J. Jang, J.
Mater. Chem. A, 2014, 2, 1642916433.
86 H.-S. Kim, I. Mora-Sero, V. Gonzalez-Pedro, F. FabregatSantiago, E. J. Juarez-Perez, N.-G. Park and J. Bisquert,
Nat. Commun., 2013, 4, DOI: 10.1038/ncomms3242.
87 E. Edri, S. Kirmayer, S. Mukhopadhyay, K. Gartsman,
G. Hodes and D. Cahen, Nat. Commun., 2014, 5, DOI:
10.1038/ncomms4461.
88 Y. Shirota and H. Kageyama, Chem. Rev., 2007, 107, 953
1010.
J. Mater. Chem. A, 2015, 3, 89438969 | 8967
View Article Online
89 J. Burschka, A. Dualeh, F. Kessler, E. Barano, N.-L. CeveyHa, C. Yi, M. K. Nazeeruddin and M. Gr

atzel, J. Am. Chem.
Soc., 2011, 133, 1804218045.
90 J. A. Christians, R. C. M. Fung and P. V. Kamat, J. Am. Chem.
Soc., 2013, 136, 758764.
91 P. Qin, S. Tanaka, S. Ito, N. Tetreault, K. Manabe,
H. Nishino, M. K. Nazeeruddin and M. Gr
atzel, Nat.
Commun., 2014, 5, DOI: 10.1038/ncomms4834.
92 J. H. Heo, S. H. Im, J. H. Noh, T. N. Mandal, C.-S. Lim,
J. A. Chang, Y. H. Lee, H.-J. Kim, A. Sarkar,
M. K. Nazeeruddin, M. Gr
atzel and S. I. Seok, Nat.
Photonics, 2013, 7, 486491.
93 M. A. Green, K. Emery, Y. Hishikawa, W. Warta and
E. D. Dunlop, Prog. Photovoltaics, 2014, 22, 701710.
94 H. Chen, X. Pan, W. Liu, M. Cai, D. Kou, Z. Huo, X. Fang and
S. Dai, Chem. Commun., 2013, 49, 72777279.
95 S. N. Habisreutinger, T. Leijtens, G. E. Eperon, S. D. Stranks,
R. J. Nicholas and H. J. Snaith, Nano Lett., 2014, 14, 5561
5568.
96 Y. Xiao, G. Han, Y. Chang, H. Zhou, M. Li and Y. Li, J. Power
Sources, 2014, 267, 18.
97 B. Cai, Y. Xing, Z. Yang, W.-H. Zhang and J. Qiu, Energy
Environ. Sci., 2013, 6, 1480.
98 E. Edri, S. Kirmayer, M. Kulbak, G. Hodes and D. Cahen, J.
Phys. Chem. Lett., 2014, 5, 429433.
99 S. Ryu, J. H. Noh, N. J. Jeon, Y. Chan Kim, W. S. Yang, J. Seo
and S. I. Seok, Energy Environ. Sci., 2014, 7, 26142618.
100 Y. S. Kwon, J. Lim, H.-J. Yun, Y.-H. Kim and T. Park, Energy
Environ. Sci., 2014, 7, 1454.
101 N. J. Jeon, J. Lee, J. H. Noh, M. K. Nazeeruddin, M. Gr
atzel
and S. I. Seok, J. Am. Chem. Soc., 2013, 135, 1908719090.
102 H. Li, K. Fu, A. Hagfeldt, M. Gr
atzel, S. G. Mhaisalkar and
A. C. Grimsdale, Angew. Chem., Int. Ed., 2014, 53, 4085
4088.
103 T. Krishnamoorthy, F. Kunwu, P. P. Boix, H. Li, T. M. Koh,
W. L. Leong, S. Powar, A. Grimsdale, M. Gr
atzel,
N. Mathews and S. G. Mhaisalkar, J. Mater. Chem. A, 2014,
2, 63056309.
104 A. Krishna, D. Sabba, H. Li, J. Yin, P. P. Boix, C. Soci,
S. G. Mhaisalkar and A. C. Grimsdale, Chem. Sci., 2014, 5,
2702.
105 N. J. Jeon, H. G. Lee, Y. C. Kim, J. Seo, J. H. Noh, J. Lee and
S. I. Seok, J. Am. Chem. Soc., 2014, 136, 78377840.
106 H. Choi, S. Paek, N. Lim, Y. H. Lee, M. K. Nazeeruddin and
J. Ko, Chem.Eur. J., 2014, 20, 1089410899.
107 J. Wang, S. Wang, X. Li, L. Zhu, Q. Meng, Y. Xiao and D. Li,
Chem. Commun., 2014, 50, 58295832.
108 Q. Chen, H. Zhou, T.-B. Song, S. Luo, Z. Hong, H.-S. Duan,
L. Dou, Y. Liu and Y. Yang, Nano Lett., 2014, 14, 41584163.
109 P. Qin, S. Paek, M. I. Dar, N. Pellet, J. Ko, M. Gr
atzel and
M. K. Nazeeruddin, J. Am. Chem. Soc., 2014, 136, 85168519.
110 D. Bi, L. Yang, G. Boschloo, A. Hagfeldt and
E. M. J. Johansson, J. Phys. Chem. Lett., 2013, 4, 15321536.
111 E. J. Juarez-Perez, M. Wubler, F. Fabregat-Santiago,
K. Lakus-Wollny, E. Mankel, T. Mayer, W. Jaegermann
and I. Mora-Sero, J. Phys. Chem. Lett., 2014, 5, 680685.
8968 | J. Mater. Chem. A, 2015, 3, 89438969
Review
112 L. Etgar, P. Gao, Z. Xue, Q. Peng, A. K. Chandiran, B. Liu,

M. K. Nazeeruddin and M. Gr
atzel, J. Am. Chem. Soc.,
2012, 134, 1739617399.
113 W. A. Laban and L. Etgar, Energy Environ. Sci., 2013, 6, 3249.
114 S. Aharon, S. Gamliel, B. E. Cohen and L. Etgar, Phys. Chem.
Chem. Phys., 2014, 16, 1051210518.
115 J. Shi, J. Dong, S. Lv, Y. Xu, L. Zhu, J. Xiao, X. Xu, H. Wu,
D. Li, Y. Luo and Q. Meng, Appl. Phys. Lett., 2014, 104,
063901.
116 F. Hao, C. C. Stoumpos, Z. Liu, R. P. Chang and
M. G. Kanatzidis, J. Am. Chem. Soc., 2014, 136, 16411
16419.
117 B.-E. Cohen, S. Gamliel and L. Etgar, APL Mater., 2014, 2,
081502.
118 S. Aharon, B. E. Cohen and L. Etgar, J. Phys. Chem. C, 2014,
118, 1716017165.
119 Z. Ku, Y. Rong, M. Xu, T. Liu and H. Han, Sci. Rep., 2013, 3,
3132.
120 A. Mei, X. Li, L. Liu, Z. Ku, T. Liu, Y. Rong, M. Xu, M. Hu,
J. Chen, Y. Yang, M. Gr
atzel and H. Han, Science, 2014,
345, 295298.
121 M. Xu, Y. Rong, Z. Ku, A. Mei, T. Liu, L. Zhang, X. Li and
H. Han, J. Mater. Chem. A, 2014, 2, 8607.
122 Z. Wei, K. Yan, H. Chen, Y. Yi, T. Zhang, X. Long, J. Li,
L. Zhang, J. Wang and S. Yang, Energy Environ. Sci., 2014,
7, 33263333.
123 E. Edri, S. Kirmayer, A. Henning, S. Mukhopadhyay,
K. Gartsman, Y. Rosenwaks, G. Hodes and D. Cahen,
Nano Lett., 2014, 14, 10001004.
124 J.-W. Lee, T.-Y. Lee, P. J. Yoo, M. Gr
atzel, S. Mhaisalkar and
N.-G. Park, J. Mater. Chem. A, 2014, 2, 92519259.
125 M. I. Dar, F. J. Ramos, Z. Xue, B. Liu, S. Ahmad,
S. A. Shivashankar, M. K. Nazeeruddin and M. Gr
atzel,
Chem. Mater., 2014, 26, 46754678.
126 B. Conings, L. Baeten, T. Jacobs, R. Dera, J. D'Haen,
J. Manca and H.-G. Boyen, APL Mater., 2014, 2, 081505.
127 A. Yella, L.-P. Heiniger, P. Gao, M. K. Nazeeruddin and
M. Gr
atzel, Nano Lett., 2014, 14, 25912596.
128 Q. Zhang, C. S. Dandeneau, X. Zhou and G. Cao, Adv.
Mater., 2009, 21, 40874108.
129 L. Wang, W. Fu, Z. Gu, C. Fan, X. Yang, H. Li and H. Chen, J.
Mater. Chem. C, 2014, 2, 90879090.
130 B. Conings, L. Baeten, C. De Dobbelaere, J. D'Haen,
J. Manca and H.-G. Boyen, Adv. Mater., 2014, 26, 20412046.
131 Y. Guo, C. Liu, K. Inoue, K. Harano, H. Tanaka and
E. Nakamura, J. Mater. Chem. A, 2014, 2, 1382713830.
132 S. Chavhan, O. Miguel, H.-J. Grande, V. Gonzalez-Pedro,
R. S. Sanchez, E. M. Barea, I. Mora-Sero and R. TenaZaera, J. Mater. Chem. A, 2014, 2, 1275412760.
133 L. Zheng, Y.-H. Chung, Y. Ma, L. Zhang, L. Xiao, Z. Chen,
S. Wang, B. Qu and Q. Gong, Chem. Commun., 2014, 50,
1119611199.
134 J. Kim, G. Kim, T. K. Kim, S. Kwon, H. Back, J. Lee, S. H. Lee,
H. Kang and K. Lee, J. Mater. Chem. A, 2014, 2, 17291
17296.
135 G. Li, R. Zhu and Y. Yang, Nat. Photonics, 2012, 6, 153161.
View Article Online
Review
136 K.-G. Lim, H.-B. Kim, J. Jeong, H. Kim, J. Y. Kim and

T.-W. Lee, Adv. Mater., 2014, 26, 64616466.
137 W. Yan, Y. Li, W. Sun, H. Peng, S. Ye, Z. Liu, Z. Bian and
C. Huang, RSC Adv., 2014, 4, 3303933046.
138 L. E. Polander, P. Pahner, M. Schwarze, M. Saalfrank,
C. Koerner and K. Leo, APL Mater., 2014, 2, 081503.
139 Z. Wu, S. Bai, J. Xiang, Z. Yuan, Y. Yang, W. Cui, X. Gao,
Z. Liu, Y. Jin and B. Sun, Nanoscale, 2014, 6, 1050510510.
140 P. Docampo, J. M. Ball, M. Darwich, G. E. Eperon and
H. J. Snaith, Nat. Commun., 2013, 4.
141 A. S. Subbiah, A. Halder, S. Ghosh, N. Mahuli, G. Hodes and
S. K. Sarkar, J. Phys. Chem. Lett., 2014, 5, 17481753.
142 J.-Y. Jeng, K.-C. Chen, T.-Y. Chiang, P.-Y. Lin, T.-D. Tsai,
Y.-C. Chang, T.-F. Guo, P. Chen, T.-C. Wen and Y.-J. Hsu,
Adv. Mater., 2014, 26, 41074113.
143 Z. Zhu, Y. Bai, T. Zhang, Z. Liu, X. Long, Z. Wei, Z. Wang,
L. Zhang, J. Wang, F. Yan and S. Yang, Angew. Chem., Int.
Ed., 2014, 53, 1257112575.
144 J. Seo, S. Park, Y. Chan Kim, N. J. Jeon, J. H. Noh, S. C. Yoon
and S. I. Seok, Energy Environ. Sci., 2014, 7, 26422646.
145 O. Malinkiewicz, C. Rold
an-Carmona, A. Soriano,
E. Bandiello, L. Camacho, M. K. Nazeeruddin and
H. J. Bolink, Adv. Energy Mater., 2014, 4, 1400345.
146 J. You, Z. Hong, Y. Yang, Q. Chen, M. Cai, T.-B. Song,
C.-C. Chen, S. Lu, Y. Liu, H. Zhou and Y. Yang, ACS Nano,
2014, 8, 16741680.
147 O. Malinkiewicz, A. Yella, Y. H. Lee, G. M. Espallargas,
M. Graetzel, M. K. Nazeeruddin and H. J. Bolink, Nat.
Photonics, 2014, 8, 128132.
148 C. Momblona, O. Malinkiewicz, C. Rold
an-Carmona,
A. Soriano, L. Gil-Escrig, E. Bandiello, M. Scheepers,
E. Edri and H. J. Bolink, APL Mater., 2014, 2, 081504.
149 Y.-Y. Yu, R.-S. Chiang, H.-L. Hsu, C.-C. Yang and C.-P. Chen,
Nanoscale, 2014, 6, 1140311410.
150 A. T. Barrows, A. J. Pearson, C. K. Kwak, A. D. F. Dunbar,
A. R. Buckley and D. G. Lidzey, Energy Environ. Sci., 2014,
7, 29442950.
151 H. Zhang, H. Azimi, Y. Hou, T. Ameri, T. Przybilla,
E. Spiecker, M. Kra, U. Scherf and C. J. Brabec, Chem.
Mater., 2014, 26, 51905193.
152 D. Liu Xiao, S.-L. Zhao, Z. Xu, F.-J. Zhang, T.-H. Zhang,
W. Gong, G. Yan, C. Kong, Y.-S. Wang and X.-R. Xu, Chin.
Phys. B, 2011, 20, 068801.
153 T. Leijtens, S. D. Stranks, G. E. Eperon, R. Lindblad,
E. M. J. Johansson, I. J. McPherson, H. Rensmo,
J. M. Ball, M. M. Lee and H. J. Snaith, ACS Nano, 2014, 8,
71477155.
154 V. D'Innocenzo, G. Grancini, M. J. P. Alcocer,
A. R. S. Kandada, S. D. Stranks, M. M. Lee, G. Lanzani,
H. J. Snaith and A. Petrozza, Nat. Commun., 2014, 5, DOI:
10.1038/ncomms4586.
155 L. Wang, C. McCleese, A. Kovalsky, Y. Zhao and C. Burda, J.
Am. Chem. Soc., 2014, 136, 1220512208.
156 H. Oga, A. Saeki, Y. Ogomi, S. Hayase and S. Seki, J. Am.
Chem. Soc., 2014, 136, 1381813825.
157 D. Nanova, A. K. Kast, M. Pfannm
oller, C. M
uller, L. Veith,
I. Wacker, M. Agari, W. Hermes, P. Erk, W. Kowalsky,
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
R. R. Schr
oder and R. Lovrincic, Nano Lett., 2014, 14,
27352740.
B. A. Gregg, J. Phys. Chem. B, 2003, 107, 46884698.
J. S. Manser and P. V. Kamat, Nat. Photonics, 2014, 8, 737
743.
J. Even, L. Pedesseau and C. Katan, J. Phys. Chem. C, 2014,
118, 1156611572.
M. D. McGehee, Nat. Mater., 2014, 13, 845846.
H.-S. Kim and N.-G. Park, J. Phys. Chem. Lett., 2014, 5, 2927
2934.
H. J. Snaith, A. Abate, J. M. Ball, G. E. Eperon, T. Leijtens,
N. K. Noel, S. D. Stranks, J. T.-W. Wang,
K. Wojciechowski and W. Zhang, J. Phys. Chem. Lett.,
2014, 5, 15111515.
E. L. Unger, E. T. Hoke, C. D. Bailie, W. H. Nguyen,
A. R. Bowring, T. Heumuller, M. G. Christoforo and
M. D. McGehee, Energy Environ. Sci., 2014, 7, 36903698.
Y. Zhao, A. Nardes and K. Zhu, Faraday Discuss., 2014, DOI:
10.1039/c4fd00128a.
T. Supasai, N. Rujisamphan, K. Ullrich, A. Chemseddine
and T. Dittrich, Appl. Phys. Lett., 2013, 103, 183906.
V. Somsongkul, F. Lang, A. R. Jeong, M. Rusu,
M. Arunchaiya and T. Dittrich, Phys. Status Solidi RRL,
2014, 08, 763766.
G. Niu, W. Li, F. Meng, L. Wang, H. Dong and Y. Qiu, J.
Mater. Chem. A, 2014, 2, 705710.
M. Gr
atzel, Nat. Mater., 2014, 13, 838842.
C. Law, L. Miseikis, S. Dimitrov, P. Shakya-Tuladhar, X. Li,
P. R. F. Barnes, J. Durrant and B. C. O'Regan, Adv. Mater.,
2014, 26, 62686273.
T. Leijtens, G. E. Eperon, S. Pathak, A. Abate, M. M. Lee and
H. J. Snaith, Nat. Commun., 2013, 4, DOI: 10.1038/
ncomms3885.
L. Badia, E. Mas-Marz
a, R. S. S
anchez, E. M. Barea,
J. Bisquert and I. Mora-Ser
o, APL Mater., 2014, 2, 081507.
M. Jorgensen, K. Norrman, S. A. Gevorgyan, T. Tromholt,
B. Andreasen and F. C. Krebs, Adv. Mater., 2012, 24, 580
612.
C. C. Stoumpos, C. D. Malliakas and M. G. Kanatzidis,
Inorg. Chem., 2013, 52, 90199038.
T. Baikie, Y. Fang, J. M. Kadro, M. Schreyer, F. Wei,
S. G. Mhaisalkar, M. Graetzel and T. J. White, J. Mater.
Chem. A, 2013, 1, 56285641.
A. Dualeh, P. Gao, S. I. Seok, M. K. Nazeeruddin and
M. Gr
atzel, Chem. Mater., 2014, 26, 61606164.
S. Aharon, A. Dymshits, A. Rotem and L. Etgar, J. Mater.
Chem. A, 2015, DOI: 10.1039/c4ta05149a.
J. H. Noh, S. H. Im, J. H. Heo, T. N. Mandal and S. I. Seok,
Nano Lett., 2013, 13, 17641769.
G. E. Eperon, V. M. Burlakov, A. Goriely and H. J. Snaith,
ACS Nano, 2013, 8, 591598.
V. M. Burlakov, G. E. Eperon, H. J. Snaith, S. J. Chapman
and A. Goriely, Appl. Phys. Lett., 2014, 104, 091602.
H. Choi, J. Jeong, H.-B. Kim, S. Kim, B. Walker, G.-H. Kim
and J. Y. Kim, Nano Energy, 2014, 7, 8085.
F. Zuo, S. T. Williams, P.-W. Liang, C.-C. Chueh, C.-Y. Liao
and A. K. Y. Jen, Adv. Mater., 2014, 26, 64546460.
J. Mater. Chem. A, 2015, 3, 89438969 | 8969

Materials Chemistry A: Journal of

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Materials Chemistry A: Journal of

Uploaded by

Copyright:

Available Formats

Journal of

Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.

Cite this: J. Mater. Chem. A, 2015, 3,

View Journal | View Issue

Perovskite-based solar cells: impact of morphology

Received 1st October 2014

School of Materials Science and Engineering, Nanyang Technological University, 50

Energy Research Institute@NTU (ERI@N), Nanyang Technological University,

This journal is The Royal Society of Chemistry 2015

either architecture. It is precisely the robustness of the

J. Mater. Chem. A, 2015, 3, 89438969 | 8943

View Article Online

Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.

Journal of Materials Chemistry A

During the same period, Lee et al. demonstrated a variation

8944 | J. Mater. Chem. A, 2015, 3, 89438969

a challenging problem. The euphoria is mainly because of the

In order to obtain good performance in planar perovskite

This journal is The Royal Society of Chemistry 2015

View Article Online

Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.

Journal of Materials Chemistry A

Simplied schematic of heterogeneous nucleation on solid

This journal is The Royal Society of Chemistry 2015

The morphology of an organometal halide perovskite lm, which

J. Mater. Chem. A, 2015, 3, 89438969 | 8945

View Article Online

Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.

Journal of Materials Chemistry A

straightforward to implement. Nevertheless, eciencies of

8946 | J. Mater. Chem. A, 2015, 3, 89438969

compact TiO2 with crystalline features on the length scale of

The solvent engineering approach demonstrated by Jeon et al. is

This journal is The Royal Society of Chemistry 2015

View Article Online

Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.

Journal of Materials Chemistry A

therefore, removes excess DMSO solvent and this encourages

evaporation rate, which is in turn dependent on its boiling point

Rapid solvent removal

This journal is The Royal Society of Chemistry 2015

Unlike their mesoscopic counterparts, planar perovskite

J. Mater. Chem. A, 2015, 3, 89438969 | 8947

View Article Online

Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.

Journal of Materials Chemistry A

system.5 One possible reason to explain this observation is the

8948 | J. Mater. Chem. A, 2015, 3, 89438969

subtle variations in the kinetics of the complex ion formation

This journal is The Royal Society of Chemistry 2015

View Article Online

between mixed-halide perovskite and TiO2, enhancing electron

Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.

Wang et al. observed a strong dependence of CH3NH3PbI3

The crystallization rate of perovskites can also be manipulated

Thermal annealing and ash annealing

Heat treatment performed on perovskite lms typically serves a

This journal is The Royal Society of Chemistry 2015

Journal of Materials Chemistry A

Solvent annealing has been demonstrated to be an attractive

View Article Online

Journal of Materials Chemistry A

Published on 18 December 2014. Downloaded by Sungkyunkwan University on 28/05/2015 14:30:31.